Weekly TMLR digest for Aug 31, 2025

[TMLR]

New certifications
==================

Featured Certification: MobileCLIP2: Improving Multi-Modal Reinforced Training

Fartash Faghri, Pavan Kumar Anasosalu Vasu, Cem Koc, Vaishaal Shankar, Alexander T Toshev, Oncel Tuzel, Hadi Pouransari

https://openreview.net/forum?id=WeF9zolng8

---

Reproducibility Certification: [Re] Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents

Oliver van Erven, Konstantinos Zafeirakis, Jacobus Smit, Julio Smidi, Luc Buijs

https://openreview.net/forum?id=EWWxSkUchO

---

Accepted papers
===============

Title: Using Platt’s scaling for calibration after undersampling — limitations and how to address them

Authors: Nathan Phelps, Daniel J Lizotte, Douglas G. Woolford

Abstract: When modelling data where the response is dichotomous and highly imbalanced, response-based sampling where a subset of the majority class is retained (i.e., undersampling) is often used to create more balanced training datasets prior to modelling. However, the models fit to this undersampled data, which we refer to as base models, generate predictions that are severely biased. There are several calibration methods that can be used to combat this bias, one of which is Platt’s scaling. Here, a logistic regression model is used to model the relationship between the base model’s original predictions and the response. Despite its popularity for calibrating models after undersampling, Platt’s scaling was not designed for this purpose. Our work presents what we believe is the first detailed study focused on the validity of using Platt’s scaling to calibrate models after undersampling. We show analytically, as well as via a simulation study, that Platt’s scaling should not be used for calibration after undersampling without critical thought. If Platt’s scaling would have been able to successfully calibrate the base model had it been trained on the entire dataset (i.e., without undersampling), then Platt’s scaling might be appropriate for calibration after undersampling. If this is not the case, we recommend a modified version of Platt’s scaling that fits a logistic generalized additive model to the logit of the base model’s predictions, as this method is theoretically motivated and performed relatively well across the settings considered in our study.

URL: https://openreview.net/forum?id=80b2zaeTUe

---

Title: Spurious Privacy Leakage in Neural Networks

Authors: Chenxiang Zhang, Jun Pang, Sjouke Mauw

Abstract: Neural networks trained on real-world data often exhibit biases while simultaneously being vulnerable to privacy attacks aimed at extracting sensitive information. Despite extensive research on each problem individually, their intersection remains poorly understood. In this work, we investigate the privacy impact of spurious correlation bias. We introduce _spurious privacy leakage_, a phenomenon in which spurious groups are significantly more vulnerable to privacy attacks than non-spurious groups. We observe that privacy disparity between groups increases in tasks with simpler objectives (e.g. fewer classes) due to spurious features. Counterintuitively, we demonstrate that spurious robust methods, designed to reduce spurious bias, fail to mitigate privacy disparity. Our analysis reveals that this occurs because robust methods can reduce reliance on spurious features for prediction, but do not prevent their memorization during training. Finally, we systematically compare the privacy of different model architectures trained with spurious data, demonstrating that, contrary to previous work, architectural choice can affect privacy evaluation.

URL: https://openreview.net/forum?id=tRXDCIgvTT

---

Title: ReHub: Linear Complexity Graph Transformers with Adaptive Hub-Spoke Reassignment

Authors: Tomer Borreda, Daniel Freedman, Or Litany

Abstract: We present ReHub, a novel graph transformer architecture that achieves linear complexity through an efficient reassignment technique between nodes and virtual nodes. Graph transformers have become increasingly important in graph learning for their ability to utilize long-range node communication explicitly, addressing limitations such as oversmoothing and oversquashing found in message-passing graph networks. However, their dense attention mechanism scales quadratically with the number of nodes, limiting their applicability to large-scale graphs. ReHub draws inspiration from the airline industry's hub-and-spoke model, where flights are assigned to optimize operational efficiency. In our approach, graph nodes (spokes) are dynamically reassigned to a fixed number of virtual nodes (hubs) at each model layer. Recent work, Neural Atoms (Li et al., 2024), has demonstrated impressive and consistent improvements over GNN baselines by utilizing such virtual nodes; their findings suggest that the number of hubs strongly influences performance. However, increasing the number of hubs typically raises complexity, requiring a trade-off to maintain linear complexity. Our key insight is that each node only needs to interact with a small subset of hubs to achieve linear complexity, even when the total number of hubs is large. To leverage all hubs without incurring additional computational costs, we propose a simple yet effective adaptive reassignment technique based on hub-hub similarity scores, eliminating the need for expensive node-hub computations. Our experiments on long-range graph benchmarks indicate a consistent improvement in results over the base method, Neural Atoms, while maintaining a linear complexity instead of $O(n^{3/2})$. Remarkably, our sparse model achieves performance on par with its non-sparse counterpart. Furthermore, ReHub outperforms competitive baselines and consistently ranks among the top performers across various benchmarks.

URL: https://openreview.net/forum?id=L4S54TUOQR

---

Title: Graph Personalized Federated Learning via Client Network Learning

Authors: Jiachen Zhou, Han Xie, Carl Yang

Abstract: Graph classification is a widely studied problem for applications such as molecule/protein function prediction and drug discovery. Powerful graph neural networks (GNNs) have demonstrated state-of-the-art performance for the classification of complex graphs, but training such models can require significant amounts of high-quality labeled graphs that are expensive to collect. When individual institutes do not possess sufficient graph data, federated learning (FL) becomes a handy solution for them to collaboratively obtain powerful graph models without directly sharing their own graph data. However, existing FL frameworks for graph data do not consider the realistic setting of personalized FL with heterogeneous data, where each client aims to leverage the data of certain other clients to boost its own model performance. In this work, inspired by graph structure learning, we propose to learn a dynamic client network that tracks the graph data similarity across clients to guide model sharing along FL. Specifically, we rely on the marginal parameters of local GNNs to dynamically learn the client network, and refer to a set of fundamental graph properties to guide its learning. Extensive experiments on three real-world graph datasets demonstrate the consistent effectiveness of our two major proposed modules, which also mutually verify the effectiveness of each other.

URL: https://openreview.net/forum?id=pyTTR4pxkU

---

Title: On the Low-Rank Parametrization of Reward Models for Controlled Language Generation

Authors: Sergey Troshin, Vlad Niculae, Antske Fokkens

Abstract: Language models trained on large amounts of data are known to produce inappropriate content in some cases and require careful tuning to be used in the real world. We revisit an effective and modular approach for controllability of the language models, when an external expert model guides the decoding. Particularly, we zoom in into the parametrization choice of an external expert, highlighting the difference between low-rank and higher-rank parametrizations. Higher-rank experts are designed to support high flexibility when representing the rewards, leading to higher computational costs during decoding. However, we demonstrate that they might not use their full flexibility. By analyzing the recently proposed reward-augmented decoding approach (RAD), which uses a higher-rank expert model, we introduce a simpler but more efficient low-rank parametrization of the expert model enabling fast and effective guided decoding. We empirically show that the low-rank RAD performs on par with the more flexible RAD on a detoxification and a sentiment control task, while requiring only a single reward model call per generated token.

URL: https://openreview.net/forum?id=cjRsEGLT8B

---

Title: Change Point Detection on A Separable Model for Dynamic Networks

Authors: Yik Lun Kei, Hangjian Li, Yanzhen Chen, OSCAR HERNAN MADRID PADILLA

Abstract: This paper studies the unsupervised change point detection problem in time series of networks using the Separable Temporal Exponential-family Random Graph Model (STERGM). Inherently, dynamic network patterns are complex due to dyadic and temporal dependence, and change points detection can identify the discrepancies in the underlying data generating processes to facilitate downstream analysis. In particular, the STERGM that utilizes network statistics and nodal attributes to represent the structural patterns is a flexible and parsimonious model to fit dynamic networks. We propose a new estimator derived from the Alternating Direction Method of Multipliers (ADMM) procedure and Group Fused Lasso (GFL) regularization to simultaneously detect multiple time points where the parameters of a time-heterogeneous STERGM have shifted. Experiments on both simulated and real data show good performance of the proposed framework, and an R package CPDstergm is developed to implement the method.

URL: https://openreview.net/forum?id=DSNJykzHF3

---

Title: Goal-Conditioned Data Augmentation for Offline Reinforcement Learning

Authors: Xingshuai Huang, Di Wu, Benoit Boulet

Abstract: Offline reinforcement learning (RL) enables policy learning from pre-collected offline datasets, relaxing the need to interact directly with the environment. However, limited by the quality of offline datasets, it generally fails to learn well-qualified policies in suboptimal datasets. To address datasets with insufficient optimal demonstrations, we introduce Goal-cOnditioned Data Augmentation (GODA), a novel goal-conditioned diffusion-based method for augmenting samples with higher quality. Leveraging recent advancements in generative modelling, GODA incorporates a novel return-oriented goal condition with various selection mechanisms. Specifically, we introduce a controllable scaling technique to provide enhanced return-based guidance during data sampling. GODA learns a comprehensive distribution representation of the original offline datasets while generating new data with selectively higher-return goals, thereby maximizing the utility of limited optimal demonstrations. Furthermore, we propose a novel adaptive gated conditioning method for processing noisy inputs and conditions, enhancing the capture of goal-oriented guidance. We conduct experiments on the D4RL benchmark and real-world challenges, specifically traffic signal control (TSC) tasks, to demonstrate GODA's effectiveness in enhancing data quality and superior performance compared to state-of-the-art data augmentation methods across various offline RL algorithms.

URL: https://openreview.net/forum?id=8K16dplpE0

---

Title: Rectified Robust Policy Optimization for Model-Uncertain Constrained Reinforcement Learning without Strong Duality

Authors: Shaocong Ma, Ziyi Chen, Yi Zhou, Heng Huang

Abstract: The goal of robust constrained reinforcement learning (RL) is to optimize an agent's performance under the worst-case model uncertainty while satisfying safety or resource constraints. In this paper, we demonstrate that strong duality does not generally hold in robust constrained RL, indicating that traditional primal-dual methods may fail to find optimal feasible policies. To overcome this limitation, we propose a novel primal-only algorithm called Rectified Robust Policy Optimization (RRPO), which operates directly on the primal problem without relying on dual formulations. We provide theoretical convergence guarantees under mild regularity assumptions, showing convergence to an approximately optimal feasible policy with iteration complexity matching the best-known lower bound when the uncertainty set diameter is controlled in a specific level. Empirical results in a grid-world environment validate the effectiveness of our approach, demonstrating that RRPO achieves robust and safe performance under model uncertainties while the non-robust method can violate the worst-case safety constraints.

URL: https://openreview.net/forum?id=7l63xwAgAW

---

Title: G2D2: Gradient-Guided Discrete Diffusion for Inverse Problem Solving

Authors: Naoki Murata, Chieh-Hsin Lai, Yuhta Takida, Toshimitsu Uesaka, Bac Nguyen, Stefano Ermon, Yuki Mitsufuji

Abstract: Recent literature has effectively leveraged diffusion models trained on continuous variables as priors for solving inverse problems. Notably, discrete diffusion models with discrete latent codes have shown strong performance, particularly in modalities suited for discrete compressed representations, such as image and motion generation. However, their discrete and non-differentiable nature has limited their application to inverse problems formulated in continuous spaces. This paper presents a novel method for addressing linear inverse problems by leveraging generative models based on discrete diffusion as priors. We overcome these limitations by approximating the true posterior distribution with a variational distribution constructed from categorical distributions and continuous relaxation techniques. Furthermore, we employ a star-shaped noise process to mitigate the drawbacks of traditional discrete diffusion models with absorbing states, demonstrating that our method performs comparably to continuous diffusion techniques with lower GPU memory consumption.

URL: https://openreview.net/forum?id=fj23qnVifX

---

Title: Mamba State-Space Models Are Lyapunov-Stable Learners

Authors: John Timothy Halloran, Manbir S Gulati, Paul F Roysdon

Abstract: Mamba state-space models (SSMs) have recently outperformed state-of-the-art (SOTA) Transformer large language models (LLMs) in various tasks and been widely adapted. However, a major concern for stable learning in recurrent-based deep models (such as SSMs) is the sensitivity of their recurrent dynamics. Despite widespread adaptation, the sensitivity of Mamba’s recurrent dynamics under common fine-tuning methods–e.g., mixed-precision fine-tuning (MPFT) and parameter-efficient fine-tuning (PEFT)–remains unexplored. Empirically,
we show that Mamba LLMs are extremely stable to changes introduced by combinations of MPFT and PEFT, in stark contrast to Transformer LLMs, which we demonstrate may drastically diverge from their respective full-precision counterparts under different
combinations of MPFT and PEFT (despite the near-ubiquitous adaptation of these fine-tuning frameworks for attention-based models). The demonstrated robustness of Mamba LLMs are due to their recurrent dynamics, which we prove are guaranteed to be stable using
dynamical systems theory (in particular, Lyapunov stability). We conclude by using MPFT and PEFT to novelly study Mamba LLMs’ in-context learning (ICL) abilities on natural language tasks, thus supplementing other recent work.

URL: https://openreview.net/forum?id=wzsYQYs3dO

---

Title: Cardinality Sparsity: Applications in Matrix-Matrix Multiplications and Machine Learning

Authors: Ali Mohaddes, Johannes Lederer

Abstract: High-dimensional data has become ubiquitous across the sciences but presents computational and statistical challenges. A common approach to addressing these challenges is through sparsity. In this paper, we introduce a new concept of sparsity, called cardinality sparsity. Broadly speaking, we define a tensor as sparse if it contains only a small number of unique values. We demonstrate that cardinality sparsity can improve deep learning and tensor regression both statistically and computationally. Along the way, we generalize recent statistical theories in these fields. Most importantly, we show that cardinality sparsity has a strikingly powerful application beyond high-dimensional data analysis: it can significantly speed up matrix-matrix multiplications. For instance, we demonstrate that cardinality sparsity leads to algorithms for binary-matrix multiplication that outperform state-of-the-art algorithms by a substantial margin. Additionally, another crucial aspect of this sparsity is minimizing memory usage. By executing matrix multiplication in the compressed domain, we can significantly lower the amount of memory needed to store the input data.

URL: https://openreview.net/forum?id=zoSRSpGu9C

---

Title: On Time Series Clustering with Graph Neural Networks

Authors: Jonas Berg Hansen, Andrea Cini, Filippo Maria Bianchi

Abstract: Graph clustering and pooling operators have been adopted in graph-based architectures to capture meaningful patterns in time series data by leveraging both temporal and relational structures. However, the contribution of each design choice and the behavior of different operators remain underexplored. This work introduces a streamlined deep learning framework based on a spatio-temporal graph neural network (STGNN) for clustering time series, which can leverage prior knowledge on the spatial structure of the data. The STGNN-based model flexibly identifies clusters in various data settings through an encoder-decoder architecture with a bottleneck, showing that a spatio-temporal approach can identify meaningful clusters even in datasets that do not explicitly include spatial relations. We validate the framework’s qualitative performance through experiments on synthetic and real-world data, showing its effectiveness in different scenarios. We also provide a heuristic for model selection in unsupervised settings via a self-supervised forecasting loss. Code is available at https://github.com/NGMLGroup/Time-Series-Clustering-with-GNNs

URL: https://openreview.net/forum?id=MHQXfiXsr3

---

Title: Variance Reduced Smoothed Functional REINFORCE Policy Gradient Algorithms

Authors: Shalabh Bhatnagar, Deepak H R

Abstract: We revisit the REINFORCE policy gradient algorithm from the literature that works with reward (or cost) returns obtained over episodes or trajectories. We propose a major enhancement to the basic algorithm where we estimate the policy gradient using a smoothed
functional (random perturbation) gradient estimator obtained from direct function measurements. To handle the issue of high variance that is typical of REINFORCE, we propose two independent enhancements to the basic scheme: (i) use the sign of the increment instead
of the original (full) increment that results in smoother convergence and (ii) use clipped gradient estimates as proposed in the Proximal Policy Optimization (PPO) based scheme. We prove the asymptotic convergence of all algorithms and show the results of several experiments on various MuJoCo locomotion tasks wherein we compare the performance of our algorithms with the recently proposed ARS algorithms in the literature as well as other well known algorithms namely A2C, PPO and TRPO. Our algorithms are seen to be competitive
against all algorithms and in fact show the best results on a majority of experiments.

URL: https://openreview.net/forum?id=yagxqSJbiY

---

Title: Streaming Heteroscedastic Probabilistic PCA with Missing Data

Authors: Kyle Gilman, David Hong, Jeffrey A Fessler, Laura Balzano

Abstract: Streaming principal component analysis (PCA) is an integral tool in large-scale machine learning for rapidly estimating low-dimensional subspaces from very high-dimensional data arriving at a high rate. However, modern datasets increasingly combine data from a variety of sources, and thus may exhibit heterogeneous quality across samples. Standard streaming PCA algorithms do not account for non-uniform noise, so their subspace estimates can quickly degrade. While the recently proposed Heteroscedastic Probabilistic PCA Technique (HePPCAT) addresses this heterogeneity, it was not designed to handle streaming data that may exhibit non-stationary behavior. Moreover, HePPCAT does not allow for missing entries in the data, which can be common in streaming data. This paper proposes the Streaming HeteroscedASTic Algorithm for PCA (SHASTA-PCA) to bridge this divide. SHASTA-PCA employs a stochastic alternating expectation-maximization approach that jointly learns the low-rank latent factors and the unknown noise variances from streaming data that may have missing entries and heteroscedastic noise, all while maintaining a low memory and computational footprint. Numerical experiments demonstrate the superior subspace estimation of our method compared to state-of-the-art streaming PCA algorithms in the heteroscedastic setting. Finally, we illustrate SHASTA-PCA applied to highly heterogeneous real data from astronomy.

URL: https://openreview.net/forum?id=lb2rPLuP9X

---

Title: Fast and Cost-effective Speculative Edge-Cloud Decoding with Early Exits

Authors: Yeshwanth Venkatesha, Souvik Kundu, Priyadarshini Panda

Abstract: Large Language Models (LLMs) enable various applications on edge devices such as smartphones, wearables, and embodied robots. However, their deployment often depends on expensive cloud-based APIs, creating high operational costs, which limit access for smaller organizations and raise sustainability concerns. Certain LLMs can be deployed on-device, offering a cost-effective solution with reduced latency and improved privacy. Yet, limited computing resources constrain the size and accuracy of models that can be deployed, necessitating a collaborative design between edge and cloud. We propose a fast and cost-effective speculative edge-cloud decoding framework with a large target model on the server and a small draft model on the device. By introducing early exits in the target model, tokens are generated mid-verification, allowing the client to preemptively draft subsequent tokens before final verification, thus utilizing idle time and enhancing parallelism between edge and cloud. Using an NVIDIA Jetson Nano (client) and an A100 GPU (server) with Vicuna-68M (draft) and Llama2-7B (target) models, our method achieves up to a 35% reduction in latency compared to cloud-based autoregressive decoding, with an additional 11% improvement from preemptive drafting. To demonstrate real-world applicability, we deploy our method on the Unitree Go2 quadruped robot using Vision-Language Model (VLM) based control, achieving a 21% speedup over traditional cloud-based autoregressive decoding. These results demonstrate the potential of our framework for real-time LLM and VLM applications on resource-constrained edge devices.

URL: https://openreview.net/forum?id=PTIUjARnbc

---

Title: TFAR: A Training-Free Framework for Autonomous Reliable Reasoning in Visual Question Answering

Authors: Zhuo Zhi, Chen Feng, Adam Daneshmend, Mine Orlu, Andreas Demosthenous, Lu Yin, Da Li, Ziquan Liu, Miguel R. D. Rodrigues

Abstract: Recent approaches introduce chain-of-thought (CoT) reasoning to mitigate the challenges, such as hallucination and reasoning deficit in multimodal large language models (MLLMs) and enhance performance. However, existing CoT-based methods often rely on extensive data annotation and training. To overcome these limitations, we propose a training-free framework for autonomous and reliable reasoning (TFAR), which only uses common lightweight vision tools to improve the reasoning ability of MLLMs. TFAR enables an MLLM to autonomously and accurately identify relevant regions of interest (RoIs) and support CoT reasoning, without requiring additional training or annotations, and with low computational overhead during inference. However, the use of external tools will introduce noise and uncertainty. To mitigate the uncertainty introduced by external tools and select the optimal pathway, we propose a conformal prediction-based uncertainty quantification method that calibrates the outputs from external tools and dynamically selects the most appropriate tool based on the MLLM’s output uncertainty. Experiments across five datasets demonstrate that TFAR improves performance over the base MLLM by an average of 4.6$\%$, in some cases even outperforming fine-tuned baselines, while maintaining low inference cost. These results offer new insights into training-free CoT guidance for MLLMs and underscore the value of reliable visual tools.

URL: https://openreview.net/forum?id=cBAKeZN3jy

---

Title: nnActive: A Framework for Evaluation of Active Learning in 3D Biomedical Segmentation

Authors: Carsten T. Lüth, Jeremias Traub, Kim-Celine Kahl, Till J. Bungert, Lukas Klein, Lars Krämer, Paul F Jaeger, Fabian Isensee, Klaus Maier-Hein

Abstract: Semantic segmentation is crucial for various biomedical applications, yet its reliance on large annotated datasets presents a significant bottleneck due to the high cost and specialized expertise required for manual labeling. Active Learning (AL) aims to mitigate this challenge by selectively querying the most informative samples, thereby reducing annotation effort.
However, in the domain of 3D biomedical imaging, there remains no consensus on whether AL consistently outperforms Random sampling strategies. Current methodological assessment is hindered by the wide-spread occurrence of four pitfalls with respect to AL method evaluation. These are (1) restriction to too few datasets and annotation budgets, (2) training 2D models on 3D images and not incorporating partial annotations, (3) Random baseline not being adapted to the task, and (4) measuring annotation cost only in voxels.
In this work, we introduce nnActive, an open-source AL framework that systematically overcomes the aforementioned pitfalls by (1) means of a large scale study evaluating 8 Query Methods on four biomedical imaging datasets and three label regimes, accompanied by four large-scale ablation studies, (2) extending the state-of-the-art 3D medical segmentation method nnU-Net by using partial annotations for training with 3D patch-based query selection, (3) proposing Foreground Aware Random sampling strategies tackling the foreground-background class imbalance commonly encountered in 3D medical images and (4) propose the foreground efficiency metric, which captures that the annotation cost for background- compared to foreground-regions is very low. We reveal the following key findings: (A) while all AL methods outperform standard Random sampling, none reliably surpasses an improved Foreground Aware Random sampling; (B) the benefits of AL depend on task specific parameters like number of classes and their locations; (C) Predictive Entropy is overall the best performing AL method, but likely requires the most annotation effort; (D) AL performance can be improved with more compute intensive design choices like longer training and smaller query sizes. As a holistic, open-source framework, nnActive has the potential to act as a catalyst for research and application of AL in 3D biomedical imaging. Code is at: \href{https://github.com/MIC-DKFZ/nnActive}{https://github.com/MIC-DKFZ/nnActive}

URL: https://openreview.net/forum?id=AJAnmRLJjJ

---

Title: On the Convergence Rates of Federated Q-Learning across Heterogeneous Environments

Authors: Muxing Wang, Pengkun Yang, Lili Su

Abstract: Large-scale multi-agent systems are often deployed across wide geographic areas, where agents interact with heterogeneous environments. There is an emerging interest in understanding the role of heterogeneity in the performance of the federated versions of classic reinforcement learning algorithms. In this paper, we study synchronous federated Q-learning, which aims to learn an optimal Q-function by having $K$ agents average their local Q-estimates per $E$ iterations. We provide a fine-grained characterization of the error evolution, which decays to zero as the number of iterations $T$ increases. When $K(E-1)$ is below a certain threshold, similar to the homogeneous environment settings, there is a linear speed-up concerning $K$. The slow convergence of having $E>1$ turns out to be fundamental rather than an artifact of our analysis. We prove that, for a wide range of stepsizes, the $\ell_{\infty}$ norm of the error cannot decay faster than $\Theta_R (\frac{E}{(1-\gamma)T})$, where $\Theta_R$ only hides numerical constants and the specific choice of reward values. In addition, our experiments demonstrate that the convergence exhibits an interesting two-phase phenomenon. For any given stepsize, there is a sharp phase transition of the convergence: the error decays rapidly in the beginning yet later bounces up and stabilizes.

URL: https://openreview.net/forum?id=EkLAG3gt3g

---

Title: MobileCLIP2: Improving Multi-Modal Reinforced Training

Authors: Fartash Faghri, Pavan Kumar Anasosalu Vasu, Cem Koc, Vaishaal Shankar, Alexander T Toshev, Oncel Tuzel, Hadi Pouransari

Abstract: Foundation image-text models such as CLIP with zero-shot capabilities enable a wide array of applications. MobileCLIP is a recent family of image-text models at 3-15ms latency and 50-150M parameters with state-of-the-art zero-shot accuracy. The main ingredients in MobileCLIP were its low-latency and light architectures and a novel multi-modal reinforced training that made knowledge distillation from multiple caption-generators and CLIP teachers efficient, scalable, and reproducible. In this paper, we improve the multi-modal reinforced training of MobileCLIP through: 1) better CLIP teacher ensembles trained on the DFN dataset, 2) improved captioner teachers trained on the DFN dataset and fine-tuned on a diverse selection of high-quality image-caption datasets. We discover new insights through ablations such as the importance of temperature tuning in contrastive knowledge distillation, the effectiveness of caption-generator fine-tuning for caption diversity, and the additive improvement from combining synthetic captions generated by multiple models. We train a new family of models called MobileCLIP2 and achieve state-of-the-art ImageNet-1k zero-shot accuracies at low latencies. In particular, we observe 2.2% improvement in ImageNet-1k accuracy for MobileCLIP2-B compared with MobileCLIP-B architecture. Notably, MobileCLIP2-S4 matches the zero-shot accuracy of SigLIP-SO400M/14 on ImageNet-1k while being 2× smaller and improves on DFN ViT-L/14 at 2.5× lower latency. We release our pretrained models and the data generation code. The data generation code makes it easy to create new reinforced datasets with arbitrary teachers using distributed scalable processing.

URL: https://openreview.net/forum?id=WeF9zolng8

---

Title: [Re] Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents

Authors: Oliver van Erven, Konstantinos Zafeirakis, Jacobus Smit, Julio Smidi, Luc Buijs

Abstract: Large Language Models (LLMs) are increasingly used in strategic decision-making environments, including game-theoretic scenarios where multiple agents interact under predefined rules. One such setting is the common pool resource environment. In this study, we build upon Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents (Piatti et al., 2024), a framework designed to test cooperation strategies among LLM agents. We begin by replicating their results to a large degree to validate the framework, reproducing the original claims regarding model scale in their simulation environment. Then, we extend the analysis to include models that represent the recent reasoning paradigm: Phi-4, DeepSeek-R1, and one of the distilled variants, which show improvements over their baseline counterparts but come at a higher computational cost. Here, we identify a notable trend: specialized models with reasoning-oriented training outperform general-purpose models of similar scale in this environment. Finally, we investigate the impact of different experiments, including the veil of ignorance mechanism and other prompting strategies based on universalization principles with varying levels of abstraction. Our results suggest that older models benefit significantly from explicit boundary conditions, whereas newer models demonstrate greater robustness to implicit constraints.

URL: https://openreview.net/forum?id=EWWxSkUchO

---

Title: Rollout Total Correlation for Deep Reinforcement Learning

Authors: Bang You, Huaping Liu, Jan Peters, Oleg Arenz

Abstract: Learning task-relevant representations is crucial for reinforcement learning. Recent approaches aim to learn such representations by improving the temporal consistency in the observed transitions. However, they only consider individual transitions and can fail to achieve long-term consistency. Instead, we argue that capturing aspects of the state that correlate with other states and actions of the trajectory---even more distant in the future---could further help in extracting task-relevant information. Hence, in this paper we investigate how to learn representations by maximizing the rollout total correlation, the correlation among all learned representations and actions within the trajectories produced by the agent. For improving rollout total correlation, we propose to combine two complementary lower bounds based on a generative and a discriminative model, combined with a simple and effective technique of chunk-wise mini-batching. Furthermore, we propose an intrinsic reward based on the learned representation for better exploration. Experimental evaluations on a set of challenging image-based simulated control tasks show that our method achieves better sample efficiency, and robustness to both white noise and natural video backgrounds compared to leading baselines.

URL: https://openreview.net/forum?id=qTdRJAL8Li

---

Title: Mixture of Balanced Information Bottlenecks for Long-Tailed Visual Recognition

Authors: Yifan Lan, Cai xin, Jun Cheng, Shan Tan

Abstract: Deep neural networks (DNNs) have achieved significant success in various applications with large-scale and balanced data. However, data in real-world visual recognition are usually long-tailed, bringing challenges to efficient training and deployment of DNNs. Information bottleneck (IB) is an elegant approach for representation learning. In this paper, we propose a balanced information bottleneck (BIB) approach, in which loss function re-balancing and self-distillation techniques are integrated into the original IB network. BIB is thus capable of learning a sufficient representation with essential label-related information fully preserved for long-tailed visual recognition. To further enhance the representation learning capability, we also propose a novel structure of mixture of multiple balanced information bottlenecks (MBIB), where different BIBs are responsible for combining knowledge from different network layers. MBIB facilitates an end-to-end learning strategy that trains representation and classification simultaneously from an information theory perspective. We conduct experiments on commonly used long-tailed datasets, including CIFAR100-LT, ImageNet-LT, and iNaturalist 2018. Both BIB and MBIB reach state-of-the-art performance for long-tailed visual recognition.

URL: https://openreview.net/forum?id=9eiALSuZGA

---

Title: Can Masked Autoencoders Also Listen to Birds?

Authors: Lukas Rauch, René Heinrich, Ilyass Moummad, Alexis Joly, Bernhard Sick, Christoph Scholz

Abstract: Masked Autoencoders (MAEs) learn rich representations in audio classification through an efficient self-supervised reconstruction task. Yet, general-purpose models struggle in fine-grained audio domains such as bird sound classification, which demands distinguishing subtle inter-species differences under high intra-species variability. We show that bridging this domain gap requires full-pipeline adaptation beyond domain-specific pretraining data. Using BirdSet, a large-scale bioacoustic benchmark, we systematically adapt pretraining, fine-tuning, and frozen feature utilization. Our Bird-MAE sets new state-of-the-art results on BirdSet’s multi-label classification benchmark. Additionally, we introduce the parameter-efficient prototypical probing, which boosts the utility of frozen MAE features by achieving up to 37 mAP points over linear probes and narrowing the gap to fine-tuning in low-resource settings. Bird-MAE also exhibits strong few-shot generalization with prototypical probes on our newly established few-shot benchmark on BirdSet, underscoring the importance of tailored self-supervised learning pipelines for fine-grained audio domains.

URL: https://openreview.net/forum?id=GIBWR0Xo2J

---

Title: Factor Learning Portfolio Optimization Informed by Continuous-Time Finance Models

Authors: Sinong Geng, houssam nassif, Zhaobin Kuang, Anders Max Reppen, K. Ronnie Sircar

Abstract: We study financial portfolio optimization in the presence of unknown and uncontrolled system variables referred to as stochastic factors. Existing work falls into two distinct categories: (i) reinforcement learning employs end-to-end policy learning with flexible factor representation, but does not precisely model the dynamics of asset prices or factors; (ii) continuous-time finance methods, in contrast, take advantage of explicitly modeled dynamics but pre-specify, rather than learn, factor representation. We propose FaLPO (factor learning portfolio optimization), a framework that interpolates between these two approaches. Specifically, FaLPO hinges on deep policy gradient to learn a performant investment policy that takes advantage of flexible representation for stochastic factors. Meanwhile, FaLPO also incorporates continuous-time finance models when modeling the dynamics. It uses the optimal policy functional form derived from such models and optimizes an objective that combines policy learning and model calibration. We prove the convergence of FaLPO and provide performance guarantees via a finite-sample bound. On both synthetic and real-world portfolio optimization tasks, we observe that FaLPO outperforms five leading methods. Finally, we show that FaLPO can be extended to other decision-making problems with stochastic factors.

URL: https://openreview.net/forum?id=KLOJUGusVE

---

Title: Private Regression via Data-Dependent Sufficient Statistic Perturbation

Authors: Cecilia Ferrando, Daniel Sheldon

Abstract: Sufficient statistic perturbation (SSP) is a widely used method for differentially private linear regression. SSP adopts a data-independent approach where privacy noise from a simple distribution is added to sufficient statistics. However, sufficient statistics can often be expressed as linear queries and better approximated by data-dependent mechanisms. In this paper we introduce data-dependent SSP for linear regression based on post-processing privately released marginals, and find that it outperforms state-of-the-art data-independent SSP. We extend this result to logistic regression by developing an approximate objective that can be expressed in terms of sufficient statistics, resulting in a novel and highly competitive SSP approach for logistic regression. We also make a connection to synthetic data for machine learning: for models with sufficient statistics, training on synthetic data corresponds to data-dependent SSP, with the overall utility determined by how well the mechanism answers these linear queries.

URL: https://openreview.net/forum?id=gtCfDKm9ME

---

Title: FraGNNet: A Deep Probabilistic Model for Tandem Mass Spectrum Prediction

Authors: Adamo Young, Fei Wang, David Wishart, BO WANG, Russell Greiner, Hannes Rost

Abstract: Compound identification from tandem mass spectrometry (MS/MS) data is a critical step in the analysis of complex mixtures. Typical solutions for the MS/MS spectrum to compound (MS2C) problem involve comparing the unknown spectrum against a library of known spectrum-molecule pairs, an approach that is limited by incomplete library coverage. Compound to MS/MS spectrum (C2MS) models can improve retrieval rates by augmenting real libraries with predicted MS/MS spectra. Unfortunately, many existing C2MS models suffer from problems with mass accuracy, generalization, or interpretability. We develop a new probabilistic method for C2MS prediction, FraGNNet, that can efficiently and accurately simulate MS/MS spectra with high mass accuracy. Our approach formulates the C2MS problem as learning a distribution over molecule fragments. FraGNNet achieves state-of-the-art performance in terms of prediction error and surpasses existing C2MS models as a tool for retrieval-based MS2C.

URL: https://openreview.net/forum?id=UsqeHx9Mbx

---

Title: A Mixture of Exemplars Approach for Efficient Out-of-Distribution Detection with Foundation Models

Authors: Evelyn Mannix, Howard Bondell

Abstract: One of the early weaknesses identified in deep neural networks trained for image classification tasks was their inability to provide low confidence predictions on out-of-distribution (OOD) data that was significantly different from the in-distribution (ID) data used to train them. Representation learning, where neural networks are trained in specific ways that improve their ability to detect OOD examples, has emerged as a promising solution. However, these approaches require long training times and can add additional overhead to detect OOD examples. Recent developments in Vision Transformer (ViT) foundation models—large networks trained on large and diverse datasets with self-supervised approaches—also show strong performance in OOD detection, and could address these challenges. This paper presents Mixture of Exemplars (MoLAR), an efficient approach to tackling OOD detection challenges that is designed to maximise the benefit of training a classifier with a high quality, frozen, pretrained foundation model backbone. MoLAR provides strong OOD detection performance when only comparing the similarity of OOD examples to the exemplars, a small set of images chosen to be representative of the dataset, leading to \mhl{significantly reduced overhead} for OOD detection inference over other methods that provide best performance when the full ID dataset is used. Extensive experiments demonstrate the improved OOD detection performance of MoLAR in comparison to comparable approaches in both supervised and semi-supervised settings, and code is available at github.com/emannix/molar-mixture-of-exemplars.

URL: https://openreview.net/forum?id=xpKqnSJtE4

---

Title: Emergent Neural Network Mechanisms for Generalization to Objects in Novel Orientations

Authors: Avi Cooper, Daniel Harari, Tomotake Sasaki, Spandan Madan, Hanspeter Pfister, Pawan Sinha, Xavier Boix

Abstract: The capability of Deep Neural Networks (DNNs) to recognize objects in orientations outside the training data distribution is not well understood. We investigate the limitations of DNNs’ generalization capacities by systematically inspecting DNNs' patterns of success and failure across out-of-distribution (OoD) orientations. We present evidence that DNNs (across architecture types, including convolutional neural networks and transformers) are capable of generalizing to objects in novel orientations, and we describe their generalization behaviors. Specifically, generalization strengthens when training the DNN with an increasing number of familiar objects, but only in orientations that involve 2D rotations of familiar orientations. We also hypothesize how this generalization behavior emerges from internal neural mechanisms – that neurons tuned to common features between familiar and unfamiliar objects enable out of distribution generalization – and present supporting data for this theory. The reproducibility of our findings across model architectures, as well as analogous prior studies on the brain, suggests that these orientation generalization behaviors, as well as the neural mechanisms that drive them, may be a feature of neural networks in general.

URL: https://openreview.net/forum?id=4wBQTZVSHU

---

Title: Transferring Reasoning Capabilities between LLMs operating via Curriculum Learning Policy

Authors: Leonardo Ranaldi, Giulia Pucci, Fabio Massimo Zanzotto

Abstract: In-context reasoning methods, exemplified by Chain-of-Thought (CoT) (et alia.,) empower the reasoning abilities of large language models (LLMs), eliciting them to solve complex reasoning tasks step-by-step. Nevertheless, the capacities to deliver robust CoT explanations arise only in models with billions of parameters, representing a barrier to entry for many users forced to operate on a smaller model scale, i.e., Small Language Models (SLMs). Even though many companies are releasing LLMs of the same family with a reduced number of parameters, these models sometimes produce misleading answers and are unable to deliver accurate step-wise reasoned answers. This paper proposes a method to transfer step-wise reasoning over SLMs by operating via Instruction-tuning (IT) on synthetic demonstrations delivered in a pedagogically motivated manner. In particular, firstly, we propose aligning step-wise reasoning capabilities via IT using Demonstrations "taught" by LLMs teacher to SLMs students. Then, we operate via Curriculum Learning, a pedagogically motivated learning method that improves the IT phase. We analyse the impact on the downstream performances of four question-answering benchmarks. The results show that SMLs can be instructed to reason via Demonstrations delivered by LLMs. We move a step further
in research: conceiving SLMs as human learners, we expose them to a CL teaching-based approach, obtaining better results on downstream performances.

URL: https://openreview.net/forum?id=zPKqyjmyEQ

---

Title: Dextr: Zero-Shot Neural Architecture Search with Singular Value Decomposition and Extrinsic Curvature

Authors: Rohan Asthana, Joschua Conrad, Maurits Ortmanns, Vasileios Belagiannis

Abstract: Zero-shot Neural Architecture Search (NAS) typically optimises the architecture search process by exploiting the network or gradient properties at initialisation through zero-cost proxies. The existing proxies often rely on labelled data, which is usually unavailable in real-world settings. Furthermore, the majority of the current methods focus either on optimising the convergence and generalisation attributes or solely on the expressivity of the network architectures. To address both limitations, we first demonstrate how channel collinearity affects the convergence and generalisation properties of a neural network. Then, by incorporating the convergence, generalisation and expressivity in one approach, we propose a zero-cost proxy that omits the requirement of labelled data for its computation. In particular, we leverage the Singular Value Decomposition (SVD) of the neural network layer features and the extrinsic curvature of the network output to design our proxy. As a result, the proposed proxy is formulated as the simplified harmonic mean of the logarithms of two key components: the sum of the inverse of the feature condition number and the extrinsic curvature of the network output. Our approach enables accurate prediction of network performance on test data using only a single label-free data sample. Our extensive evaluation includes a total of six experiments, including the Convolutional Neural Network (CNN) search space, i.e. DARTS and the Transformer search space, i.e. AutoFormer. The proposed proxy demonstrates a superior performance on multiple correlation benchmarks, including NAS-Bench-101, NAS-Bench-201, and TransNAS-Bench-101-micro; as well as on the NAS task within the DARTS and the AutoFormer search space, all while being notably efficient. The code is available at https://github.com/rohanasthana/Dextr.

URL: https://openreview.net/forum?id=X0vPof5DVh

---

Title: Differentiable Causal Discovery of Linear Non-Gaussian Acyclic Models Under Unmeasured Confounding

Authors: Yoshimitsu Morinishi, Shohei Shimizu

Abstract: We propose a score-based method that extends the framework of the linear non- Gaussian acyclic model (LiNGAM) to address the problem of causal structure estimation in the presence of unmeasured variables. Building on the method pro- posed by Bhattacharya et al. (2021), we develop a method called ABIC LiNGAM, which assumes that error terms follow a multivariate generalized normal distribu- tion and employs continuous optimization techniques to recover acyclic directed mixed graphs (ADMGs). We demonstrate that the proposed method can esti- mate causal structures, including the possibility of identifying their orientations, rather than only Markov equivalence classes, under the assumption that the data are linear and follow a multivariate generalized normal distribution. Additionally, we provide proofs of the identifiability of the parameters in ADMGs when the er- ror terms follow a multivariate generalized normal distribution. The effectiveness of the proposed method is validated through simulations and experiments using real-world data.

URL: https://openreview.net/forum?id=HR7MFlW73I

---

Title: PrivShap: A Finer-granularity Network Linearization Method for Private Inference

Authors: Xiangrui Xu, Zhenzhen Wang, Rui Ning, Chunsheng Xin, Hongyi Wu

Abstract: Private inference applies cryptographic techniques like homomorphic encryption, garble circuit and secret sharing to keep both sides privacy in a client-server setting during inference. It is often hindered by the high communication overheads, especially at non-linear activation layers such as ReLU. Hence ReLU pruning has been widely recognized as an efficient way to accelerate private inference. Existing approaches to ReLU pruning typically rely on coarse hypothesis, which assume an inverse correlation between the importance of ReLU and linear layers or shallow activation layers have less importance for universal models, to assign the budgets according to the layer while preserving the inference accuracy. However, these assumptions are based on limited empirical evidence and can fail to generalize to diverse model architectures. In this work, we introduce a finer-granularity ReLU budget assignment approach by assessing the layer-wise importance of ReLU with the Shapley value.

To address the computational burden of exact Shapley value calculation, we propose a tree-trimming algorithm for fast estimation. We provide both theoretical guarantees and empirical validation of our method. Our extensive experiments show that we achieve better efficiency and accuracy than the state-of-the-art across diverse model architectures, activation functions, and datasets. Specifically, we only need $\sim$$2.5\times$ fewer ReLU operations to achieve a similar inference accuracy and gains up to $\sim$$8.13\%$ increase on inference accuracy with similar ReLU budgets.

URL: https://openreview.net/forum?id=7TliYmJr2m

---

Title: Enhancing Plaque Segmentation in CCTA with Prompt- based Diffusion Data Augmentation

Authors: Ruan Yizhe, Xuangeng Chu, Ziteng Cui, Yusuke Kurose, JUNICHI IHO, Yoji Tokunaga, Makoto Horie, YUSAKU HAYASHI, Keisuke Nishizawa, Yasushi Koyama, Tatsuya Harada

Abstract: Coronary computed tomography angiography (CCTA) is essential for non-invasive assessment of coronary artery disease (CAD). However, accurate segmentation of atherosclerotic plaques remains challenging due to data scarcity, severe class imbalance, and significant variability between calcified and non-calcified plaques. Inspired by DiffTumor’s tumor synthesis and PromptIR’s adaptive restoration framework, we introduce PromptLesion, a prompt-conditioned diffusion model for multi-class lesion synthesis. Unlike single-class methods, our approach integrates lesion-specific prompts within the diffusion generation process, enhancing diversity and anatomical realism in synthetic data. We validate PromptLesion on a private CCTA dataset and multi-organ tumor segmentation tasks (kidney, liver, pancreas) using public datasets, achieving superior performance compared to baseline methods. Models trained with our prompt-guided synthetic augmentation significantly improve Dice Similarity Coefficient (DSC) scores for both plaque and tumor segmentation. Extensive evaluations and ablation studies confirm the effectiveness of prompt conditioning.

URL: https://openreview.net/forum?id=hbTYt8PX9n

---

Title: Head-Specific Intervention Can Induce Misaligned AI Coordination in Large Language Models

Authors: Paul Darm, Annalisa Riccardi

Abstract: Robust alignment guardrails for large language models (LLMs) are becoming increasingly important with their widespread application. In contrast to previous studies, we demonstrate that inference-time activation interventions can bypass safety alignments and effectively steer model generations towards harmful AI coordination. Our method applies fine-grained interventions at specific attention heads, which we identify by probing each head in a simple binary choice task. We then show that interventions on these heads generalise to the open-ended generation setting, effectively circumventing safety guardrails. We demonstrate that intervening on a few attention heads is more effective than intervening on full layers or supervised fine-tuning. We further show that only a few example completions are needed to compute effective steering directions, which is an advantage over classical fine-tuning. We also demonstrate applying interventions in the negative direction can prevent a common jailbreak attack. Our results suggest that, at the attention head level, activations encode fine-grained linearly separable behaviours. Practically, the approach offers a straightforward methodology to steer large language model behaviour, which could be extended to diverse domains beyond safety requiring fine-grained control over the model output.

URL: https://openreview.net/forum?id=VY0huMBr5n

---

New submissions
===============

Title: Beyond Expectations: Learning with Stochastic Dominance Made Practical

Abstract: Stochastic dominance serves as a general framework for modeling a broad spectrum of decision preferences under uncertainty, with risk aversion as one notable example, as it naturally captures the intrinsic structure of the underlying uncertainty, in contrast to simply resorting to the expectations. Despite theoretically appealing, the application of stochastic dominance in machine learning has been scarce, due to the following challenges: i), the original concept of stochastic dominance only provides a partial order, therefore, is not amenable to serve as a general optimality criterion; and ii), an efficient computational recipe remains lacking due to the continuum nature of evaluating stochastic dominance.

In this work, we make the first attempt towards establishing a general framework of learning with stochastic dominance. We first generalize the stochastic dominance concept to enable feasible comparisons between any arbitrary pair of random variables. We next develop a simple and computationally efficient approach for finding the optimal solution in terms of stochastic dominance, which can be seamlessly plugged into many learning tasks. Numerical experiments demonstrate that the proposed method achieves comparable performance as standard risk-neutral strategies and obtains better trade-offs against risk across a variety of applications including supervised learning, reinforcement learning, and portfolio optimization.

URL: https://openreview.net/forum?id=ebyPKXsweD

---

Title: Visual-Word Tokenizer: Beyond Fixed Sets of Tokens in Vision Transformers

Abstract: The cost of deploying vision transformers increasingly represents a barrier to wider industrial adoption. Existing compression techniques require additional end-to-end fine-tuning or incur a significant drawback to energy efficiency, making them ill-suited for online (real-time) inference, where a prediction is made on any new input as it comes in. We introduce the Visual Word Tokenizer (VWT), a training-free method for reducing energy costs while retaining performance. The VWT groups visual subwords (image patches) that are frequently used into visual words while infrequent ones remain intact. To do so, intra-image or inter-image statistics are leveraged to identify similar visual concepts for sequence compression. Experimentally, we demonstrate a reduction in energy consumed of up to 47%. Comparative approaches of 8-bit quantization and token merging can lead to significantly increased energy costs (up to 500% or more). Our results indicate that VWTs are well-suited for efficient online inference with a marginal compromise on performance.

URL: https://openreview.net/forum?id=YYOS1FHYG3

---

Title: Algorithms for the preordering problem and their application to the task of jointly clustering and ordering the accounts of a social network

Abstract: The NP-hard maximum value preordering problem is both a joint relaxation and a hybrid of the clique partition problem (a clustering problem) and the partial ordering problem. Toward approximate solutions and lower bounds, we introduce a linear-time 4-approximation algorithm that constructs a maximum dicut of a subgraph and define local search heuristics. Toward upper bounds, we tighten a linear program relaxation by the class of odd closed walk inequalities that define facets, as we show, of the preorder polytope. We contribute implementations of the algorithms, apply these to the task of jointly clustering and partially ordering the accounts of published social networks, and compare the output and efficiency qualitatively and quantitatively.

URL: https://openreview.net/forum?id=cBsUnv7Cb3

---

Title: InfGraND: An Influence-Guided GNN-to-MLP Knowledge Distillation

Abstract: Graph Neural Networks (GNNs) are the go-to model for graph data analysis. However, GNNs rely on two key operations—aggregation and update, which can pose challenges for low-latency inference tasks or resource-constrained scenarios. Simple Multi-Layer Perceptrons (MLPs) offer a computationally efficient alternative. Yet, training an MLP in a supervised setting often leads to suboptimal performance. Knowledge Distillation (KD) from a GNN teacher to an MLP student has emerged to bridge this gap. However, most KD methods either transfer knowledge uniformly across all nodes or rely on graph-agnostic indicators such as prediction uncertainty. We argue this overlooks a more fundamental, graph-centric inquiry: "How important is a node to the structure of the graph?" We introduce a framework, InfGraND, an Influence-guided Graph KNowledge Distillation from GNN to MLP that addresses this by identifying and prioritizing structurally influential nodes to guide the distillation process, ensuring that the MLP learns from the most critical parts of the graph. Additionally, InfGraND embeds structural awareness in MLPs through one-time multi-hop neighborhood feature pre-computation, which enriches the student MLP’s input and thus avoids inference-time overhead. Our rigorous evaluation in transductive and inductive settings across seven benchmark datasets shows InfGraND consistently outperforms prior GNN to MLP KD methods, demonstrating its practicality for numerous latency-critical applications in real-world settings.

URL: https://openreview.net/forum?id=lfzHR3YwlD

---

Title: Efficient Audiovisual Speech Processing via MUTUD: Multimodal Training and Unimodal Deployment

Abstract: Building reliable speech systems often requires combining multiple modalities, like audio and visual cues. While such multimodal solutions frequently lead to improvements in performance and may even be critical in certain cases, they come with several constraints such as increased sensory requirements, computational cost, and modality synchronization, to mention a few. These challenges constrain the direct uses of these multimodal solutions in real-world applications. In this work, we develop approaches where the learning happens with all available modalities but the deployment or inference is done with just one or reduced modalities. To do so, we propose a Multimodal Training and Unimodal Deployment (MUTUD) framework which includes a Temporally Aligned Modality feature Estimation (TAME) module that can estimate information from missing modality using modalities present during inference. This innovative approach facilitates the integration of information across different modalities, enhancing the overall inference process by leveraging the strengths of each modality to compensate for the absence of certain modalities during inference. We apply MUTUD to various audiovisual speech tasks and show that it can reduce the performance gap between the multimodal and corresponding unimodal models to a considerable extent. MUTUD can achieve this while reducing the model size and compute compared to multimodal models, in some cases by almost 80%.

URL: https://openreview.net/forum?id=5bshBY8RDf

---

Title: Interestingness First Classifiers

Abstract: Most machine learning models are designed to maximize predictive accuracy. In this work, we explore a different goal: building classifiers that are interesting. An ``interesting classifier'' is one that uses unusual or unexpected features, even if its accuracy is lower than the best possible model. For example, predicting room congestion from CO2 levels achieves near-perfect accuracy but is unsurprising. In contrast, predicting room congestion from humidity is less accurate yet more nuanced and intriguing. We introduce EUREKA, a simple framework that selects features according to their perceived interestingness. Our method leverages large language models to rank features by their interestingness and then builds interpretable classifiers using only the selected interesting features. Across several benchmark datasets, EUREKA consistently identifies features that are non-obvious yet still predictive. For example, in the Occupancy Detection dataset, our method favors humidity over CO2 levels and light intensity, producing classifiers that achieve meaningful accuracy while offering insights. In the Twin Papers dataset, our method discovers the rule that papers with a colon in the title are more likely to be cited in the future. We argue that such models can support new ways of knowledge discovery and communication, especially in settings where moderate accuracy is sufficient but novelty and interpretability are valued.

URL: https://openreview.net/forum?id=zHvIY49qp8

---

Title: Targeted Unlearning Using Perturbed Sign Gradient Methods With Applications On Medical Images

Abstract: Machine unlearning aims to remove the influence of specific training samples from a trained model without full retraining. While prior work has largely focused on privacy-motivated settings, we recast unlearning as a general-purpose tool for post-deployment model revision. Specifically, we focus on utilizing unlearning in clinical contexts where data shifts, device deprecation, and policy changes are common. To this end, we propose a bilevel optimization formulation of boundary-based unlearning that can be solved using iterative algorithms. We provide convergence guarantees when first order algorithms are used to unlearn. Our method introduces tunable loss design for controlling the forgetting–retention tradeoff and supports novel model composition strategies that merge the strengths of distinct unlearning runs. Across benchmark and real-world clinical imaging datasets, our approach outperforms baselines on both forgetting and retention metrics, including scenarios involving imaging devices and anatomical outliers. This work establishes machine unlearning as a modular, practical alternative to retraining for real-world model maintenance in clinical applications.

URL: https://openreview.net/forum?id=XE0bJg6sQN

---

Title: RLHF in an SFT Way: From Optimal Solution to Reward-Weighted Alignment

Abstract: Reinforcement Learning from Human Feedback (RLHF) is crucial for aligning Large Language Models (LLMs) with human values. However, RLHF has been continuously challenged by its high complexity in implementation and computation consumption, specifically for online sampling-based methods like Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO). Even with recent simplifications, such as Direct Preference Optimization (DPO) that designs an offline implicit reward learning objective relying on pre-collected preference datasets, the problems of over-fitting and training instability remain hindering the alignment process from the expected optimal performance. To address the existing challenges, we propose a novel simplification of RLHF from the perspective of variational inference, called **V**ariational **A**lignment with **R**e-weighting (**VAR**). Specifically, by directly minimizing the distribution gap between the learning LLM policy and the optimal solution of RLHF, we transform the alignment objective into an offline reward-driven re-weighted supervised fine-tuning (SFT) form, which only requires minor adjustment on the SFT loss to obtain noticeable improvement on training stability and effectiveness. In comprehensive evaluation benchmarks, our objective empowers LLMs to outperform offline alignments, demonstrating superior performance in both helpfulness and harmlessness metrics (avg. $\uparrow7.16\%$ than DPO). Meanwhile, when compared to online sampling methods, our method is also comparable even better while significantly reducing computational overhead and accelerating convergence speed (over $5\times$ faster than GRPO), suggesting our approach as an efficient and effective solution in bridging the gap between efficiency and performance in LLM alignment.

URL: https://openreview.net/forum?id=jewB0UhFuj

---

Title: Self-Improving LLMs with Synthetic Data Through Dynamic Noise Preference Optimization

Abstract: Although LLMs have achieved significant success, their reliance on large volumes of human-annotated data has limited their potential for further scaling. In this situation, utilizing self-generated synthetic data has become crucial for fine-tuning LLMs without extensive human annotation. However, current methods often fail to ensure consistent improvements across iterations, with performance stagnating after only minimal updates. To overcome these challenges, we introduce Dynamic Noise Preference Optimization (DNPO), which combines dynamic sample labeling for constructing preference pairs with controlled, trainable noise injection during preference optimization. Our approach effectively prevents stagnation and enables continuous improvement. In experiments with Llama-3.2-3B and Zephyr-7B, DNPO consistently outperforms existing methods across multiple benchmarks. Additionally, with Zephyr-7B, DNPO shows a significant improvement in model-generated data quality, with a 29.4% win-loss rate gap compared to the baseline in GPT-4 evaluations.

URL: https://openreview.net/forum?id=qDexGLXpef

---

Title: Disentangled Concept-Residual Models: Bridging the Interpretability–Performance Gap for Incomplete Concept Sets

Abstract: Deploying AI in high-stakes settings requires models that are not only accurate but also interpretable and amenable to human oversight. Concept Bottleneck Models (CBMs) support these goals by structuring predictions around human-understandable concepts, enabling interpretability and post-hoc human intervenability. However, CBMs rely on a ‘complete’ concept set, requiring practitioners to define and label enough concepts to match the predictive power of black-box models. To relax this requirement, prior work introduced residual connections that bypass the concept layer and recover information missing from an incomplete concept set. While effective in bridging the performance gap, these residuals can redundantly encode concept information, a phenomenon we term \textbf{concept-residual overlap}. In this work, we investigate the effects of concept-residual overlap and evaluate strategies to mitigate it. We (1) define metrics to quantify the extent of concept-residual overlap in CRMs; (2) introduce complementary metrics to evaluate how this overlap impacts interpretability, concept importance, and the effectiveness of concept-based interventions; and (3) present \textbf{Disentangled Concept-Residual Models (D-CRMs)}, a general class of CRMs designed to mitigate this issue. Within this class, we propose a novel disentanglement approach based on minimizing mutual information (MI). Using CelebA, CIFAR100, AA2, CUB, and OAI, we show that standard CRMs exhibit significant concept-residual overlap, and that reducing this overlap with MI-based D-CRMs restores key properties of CBMs, including interpretability, functional reliance on concepts, and intervention robustness, without sacrificing predictive performance.

URL: https://openreview.net/forum?id=NKgNizwDa6

---

Title: Beyond Affinity: A Benchmark of 1D, 2D, and 3D Methods Reveals Critical Trade-offs in Structure-Based Drug Design

Abstract: Currently, the field of structure-based drug design is dominated by three main types of algorithms: search-based algorithms, deep generative models, and reinforcement learning. While existing works have typically focused on comparing models within a single algorithmic category, cross-algorithm comparisons remain scarce. In this paper, to fill the gap, we establish a benchmark to evaluate the performance of fifteen models across these different algorithmic foundations by assessing the pharmaceutical properties of the generated molecules and their docking affinities and poses with specified target proteins. We highlight the unique advantages of each algorithmic approach and offer recommendations for the design of future SBDD models. We emphasize that 1D/2D ligand-centric drug design methods can be used in SBDD by treating the docking function as a black-box oracle, which is typically neglected. Our evaluation reveals distinct patterns across model categories. 3D structure-based models excel in binding affinities but show inconsistencies in chemical validity and pose quality. 1D models demonstrate reliable performance in standard molecular metrics but rarely achieve optimal binding affinities. 2D models offer balanced performance, maintaining high chemical validity while achieving moderate binding scores. Through detailed analysis across multiple protein targets, we identify key improvement areas for each model category, providing insights for researchers to combine strengths of different approaches while addressing their limitations.

URL: https://openreview.net/forum?id=gaTwx1rzCw

---

Title: Dissecting Bias in LLMs: A Mechanistic Interpretability Perspective

Abstract: Large Language Models (LLMs) are known to exhibit social, demographic, and gender biases, often as a consequence of the data on which they are trained. In this work, we adopt a mechanistic interpretability approach to analyze how such biases are structurally represented within models such as GPT-2 and Llama2. Focusing on demographic and gender biases, we explore different metrics to identify the internal edges responsible for biased behavior. We then assess the stability, localization, and generalizability of these components across dataset and linguistic variations. Through systematic ablations, we demonstrate that bias-related computations are highly localized, often concentrated in a small subset of layers. Moreover, the identified components change across fine-tuning settings, including those unrelated to bias. Finally, we show that removing these components not only reduces biased outputs but also affects other NLP tasks, such as named entity recognition and linguistic acceptability judgment because of the sharing of important components with these tasks.

URL: https://openreview.net/forum?id=EpQ2CBJTjD

---

Title: Efficient and Unbiased Sampling from Boltzmann Distributions via Variance-Tuned Diffusion Models

Abstract: Score-based diffusion models (SBDMs) are powerful amortized samplers for Boltzmann distributions; however, imperfect score estimates bias downstream Monte Carlo estimates. Classical importance sampling (IS) can correct this bias, but computing exact likelihoods requires solving the probability-flow ordinary differential equation (PF–ODE), a procedure that is prohibitively costly and scales poorly with dimensionality. We introduce Variance-Tuned Diffusion Importance Sampling (VT-DIS), a lightweight post-training method that adapts the per-step noise covariance of a pretrained SBDM by minimizing the $\alpha$-divergence $(\alpha=2)$ between its forward diffusion and reverse denoising trajectories. VT-DIS assigns a single trajectory-wise importance weight to the joint forward–reverse process, yielding unbiased expectation estimates at test time with negligible overhead compared to standard sampling. On the DW-4, LJ-13, and alanine-dipeptide benchmarks, VT-DIS achieves effective sample sizes of approximately 80%, 35%, and 3.5%, respectively, while using only a fraction of the computational budget required by vanilla diffusion + IS or PF-ODE–based IS.

URL: https://openreview.net/forum?id=Jq2dcMCS5R

---

Title: Generalization bound for a Shallow Transformer trained using Gradient Descent

Abstract: In this work, we develop a norm-based generalization bound for a shallow Transformer model trained using Gradient Descent. This is achieved in three major steps i.e., (a) Defining a class of Transformer models whose weights stay close to their initialization during training. (b) Upper bounding the Rademacher complexity of this class. (c) Upper bounding the empirical loss of all transformer models belonging to the above-defined class for all training steps. We end up with an upper bound on the true loss which tightens sublinearly with increasing number of training examples $N$ for all values of model dimension $d_m$. We also perform experiments on MNIST dataset to support our theoretical findings.

URL: https://openreview.net/forum?id=t3iUeMOT8Z

---

Title: Multi-Modal Foundation Models for Computational Pathology: A Survey

Abstract: Foundation models have emerged as a powerful paradigm in computational pathology (CPath), enabling scalable and generalizable analysis of histopathological images. While early developments centered on uni-modal models trained solely on visual data, recent advances have highlighted the promise of multi-modal foundation models that integrate heterogeneous data sources such as textual reports, structured domain knowledge, and molecular profiles. In this survey, we provide a comprehensive and up-to-date review of multi-modal foundation models in CPath, with a particular focus on models built upon hematoxylin and eosin (H&E) stained whole slide images (WSIs) and tile-level representations. We categorize 32 state-of-the-art multi-modal foundation models into three major paradigms: vision-language, vision-knowledge graph, and vision-gene expression. We further divide vision-language models into non-LLM-based and LLM-based approaches. Additionally, we analyze 28 available multi-modal datasets tailored for pathology, grouped into image-text pairs, instruction datasets, and image-other modality pairs. Our survey also presents a taxonomy of downstream tasks, highlights training and evaluation strategies, and identifies key challenges and future directions. We aim for this survey to serve as a valuable resource for researchers and practitioners working at the intersection of pathology and AI.

URL: https://openreview.net/forum?id=NZ7GSH92cY

---

Title: State Combinatorial Generalization In Decision Making With Conditional Diffusion Models

Abstract: Many real-world decision-making problems are combinatorial in nature, where states (e.g., surrounding traffic of a self-driving car) can be seen as a combination of basic elements (e.g., pedestrians, trees, and other cars). Due to combinatorial complexity, observing all combinations of basic elements in the training set is infeasible, which leads to an essential yet understudied problem of zero-shot generalization to states that are unseen combinations of previously seen elements. In this work, we first formalize this problem and then demonstrate how existing value-based reinforcement learning (RL) algorithms struggle due to unreliable value predictions in unseen states. We argue that this problem cannot be addressed with exploration alone, but requires more expressive and generalizable models. We demonstrate that behavior cloning with a conditioned diffusion model trained on successful trajectory generalizes better to states formed by new combinations of seen elements than traditional RL methods. Through experiments in maze, driving, and multiagent environments, we show that conditioned diffusion models outperform traditional RL techniques and highlight the broad applicability of our problem formulation.

URL: https://openreview.net/forum?id=XB1dd01Ozz

---

Title: Rewarding the Rare: Maverick-Aware Shapley Valuation in Federated Learning

Abstract: Federated Learning (FL) allows clients to train a model collaboratively without sharing their private data. Shapley value (SV) provides a principled way to quantify client contributions in FL. However, existing SV methods use uniform per-class weighting during validation, treating all classes as equally important. This uniform weighting breaks down in the presence of clients with underrepresented or rare classes, also referred to as Mavericks. Such clients are often undervalued due to lower model performance on these challenging classes, despite their critical role in improving generalization. To address this, we introduce a Maverick-aware Shapley valuation framework that reweights validation scores based on per-class accuracy, assigning greater importance to classes where models perform poorly. Building on this, we design FedMS, a Maverick-Shapley client selection mechanism that leverages our refined contribution scores to guide intelligent client selection. Experiments on benchmark datasets demonstrate that FedMS improves model performance and better recognizes valuable client contributions, even under scenarios involving adversaries, free-riders, and skewed or rare-class distributions.

URL: https://openreview.net/forum?id=JtybGfTUdq

---

Title: Extracting and Following Paths for Robust Relational Reasoning with Large Language Models

Abstract: Large language models (LLMs) possess vast semantic knowledge but often struggle with complex reasoning tasks, particularly in relational reasoning problems such as kinship or spatial reasoning. In this paper, we present Path-of-Thoughts (PoT), a novel framework designed to tackle relation reasoning by decomposing the task into three key stages: graph extraction, path identification, and reasoning. Unlike previous approaches, PoT efficiently extracts a task-agnostic graph that identifies crucial entities, relations, and attributes within the problem context. Subsequently, PoT identifies relevant reasoning chains within the graph corresponding to the posed question, facilitating inference of potential answers. Experimental evaluations on four benchmark datasets, demanding long reasoning chains, demonstrate that PoT surpasses state-of-the-art baselines by a significant margin (maximum 21.3%) without necessitating fine-tuning or extensive LLM calls. Furthermore, as opposed to prior neuro-symbolic methods, PoT exhibits improved resilience against LLM errors by leveraging the compositional nature of graphs.

URL: https://openreview.net/forum?id=EbELaNKmZK

---

Title: Diffusion Self-Weighted Guidance for Offline Reinforcement Learning

Abstract: Offline reinforcement learning (RL) recovers the optimal policy $\pi$ given historical observations of an agent. In practice, $\pi$ is modeled as a weighted version of the agent's behavior policy $\mu$, using a weight function $w$ working as a critic of the agent's behavior. Though recent approaches to offline RL based on diffusion models have exhibited promising results, the computation of the required scores is challenging due to their dependence on the unknown $w$. In this work, we alleviate this issue by constructing a diffusion over both the actions and the weights. With the proposed setting, the required scores are directly obtained from the diffusion model without learning extra networks. Our main conceptual contribution is a novel guidance method, where guidance (which is a function of $w$) comes from the same diffusion model, therefore, our proposal is termed Self-Weighted Guidance (SWG). We show that SWG generates samples from the desired distribution on toy examples and performs on par with state-of-the-art methods on D4RL's challenging environments, while maintaining a streamlined training pipeline. We further validate SWG through ablation studies on weight formulations and scalability.

URL: https://openreview.net/forum?id=jmXBnpmznv

---

Title: STLDM: Spatio-Temporal Latent Diffusion Model for Precipitation Nowcasting

Abstract: Precipitation nowcasting is a critical spatio-temporal prediction task for society to prevent severe damage owing to extreme weather events. Despite the advances in this field, the underlying complex and stochastic nature of this task still poses challenges to previous approaches. Specifically, deterministic models produce blurry predictions while generative models suffer from poor accuracy. In this paper, we present a simple yet effective model architecture termed STLDM, which learns the latent representation from end to end alongside both the Variational Autoencoder and the conditioning network. Experimental results across multiple radar datasets demonstrate that the proposed STLDM is more effective and superior to the state of the art.

URL: https://openreview.net/forum?id=f4oJwXn3qg

---

Title: ComFe: An Interpretable Head for Vision Transformers

Abstract: Interpretable computer vision models explain their classifications through comparing the distances between the local embeddings of an image and a set of prototypes that represent the training data. However, these approaches introduce additional hyper-parameters that need to be tuned to apply to new datasets, scale poorly, and are more computationally intensive to train in comparison to black-box approaches. In this work, we introduce Component Features (ComFe), a highly scalable interpretable-by-design image classification head for pretrained Vision Transformers (ViTs) that can obtain competitive performance in comparison to comparable non-interpretable methods. To our knowledge, ComFe is the first interpretable head and unlike other interpretable approaches can be readily applied to large-scale datasets such as ImageNet-1K. Additionally, ComFe provides improved robustness and outperforms previous interpretable approaches on key benchmark datasets while using a consistent set of hyperparameters and without finetuning the pretrained ViT backbone. With only global image labels and no segmentation or part annotations, ComFe can identify consistent component features within an image and determine which of these features are informative in making a prediction. Code is available at https://anonymous.4open.science/r/cospress-83E3/README.md.

URL: https://openreview.net/forum?id=cI4wrDYFqE

---

Title: Preserving Angles Improves Feature Distillation

Abstract: Knowledge distillation methods compress models by training a student network using the classification outputs of a high quality teacher model, but can fail to effectively transfer the properties of computer vision foundation models from the teacher to the student. While it has been recently shown that feature distillation—where a teacher model's output features are replicated instead—can reproduce performance for foundation models across numerous downstream tasks, they fall short in matching critical properties such as robustness and out-of-distribution (OOD) detection performance. This paper overcomes this shortcoming by introducing Cosine-similarity Preserving Compression (CosPress), a feature distillation technique that learns a mapping to compress the latent space of the teacher model into the smaller latent space of the student, by preserving the cosine similarities between image embeddings. This enables direct optimisation of the student network and produces a more faithful reproduction of the teacher's properties. It is shown that distillation with CosPress on a variety of datasets, including ImageNet, produces more accurate models with greater performance on generalisability, robustness and OOD detection benchmarks, and that this technique provides a competitive pathway for training highly performant lightweight models on small datasets. Code is available at https://anonymous.4open.science/r/cospress-83E3/README.md.

URL: https://openreview.net/forum?id=ZEhgODZkWU

---

Title: HopCast: Calibration of Autoregressive Dynamics Models

Abstract: Deep learning models are often trained to approximate dynamical systems that can be modeled using differential equations. Many of these models are optimized to predict one step ahead; such approaches produce calibrated one-step predictions if the predictive model can quantify uncertainty, such as Deep Ensembles. At inference time, multi-step predictions are generated via autoregression, which needs a sound uncertainty propagation method to produce calibrated multi-step predictions. This work introduces an alternative Predictor-Corrector approach named HopCast that uses Modern Hopfield Networks (MHN) to learn the errors of a deterministic Predictor that approximates the dynamical system. The Corrector predicts a set of errors for the Predictor's output based on a context state at any timestep during autoregression. The set of errors creates sharper and well-calibrated prediction intervals with higher predictive accuracy compared to baselines without uncertainty propagation. The calibration and prediction performances are evaluated across a set of dynamical systems. This work is also the first to benchmark existing uncertainty propagation methods based on calibration errors.

URL: https://openreview.net/forum?id=wsO6nxvGof

---

Title: Training Dynamics of Learning 3D-Rotational Equivariance

Abstract: While data augmentation is widely used to train symmetry-agnostic models, it remains unclear how quickly and effectively they learn to respect symmetries.
We investigate this by deriving a principled measure of equivariance error that, for convex losses, calculates the percent of total loss attributable to imperfections in learned symmetry.
We focus our empirical investigation to 3D-rotation equivariance on high-dimensional molecular tasks (flow matching, force field prediction, denoising voxels) and find that models rapidly become nearly equivariant within 1k-10k training steps, a result robust to model and dataset size.
This happens because learning 3D-rotational equivariance is an easier learning task, with a smoother and better-conditioned loss landscape, than the main prediction task.
We then theoretically characterize learning dynamics for models that are nearly equivariant, as ``stochastic equivariant learning dynamics'', via analyses that also hold beyond 3D rotations.
For 3D rotations, the loss penalty for non-equivariant models is small throughout training, so they may achieve lower test loss than equivariant models per GPU-hour unless the equivariant ``efficiency gap'' is narrowed.

URL: https://openreview.net/forum?id=DLOIAW18W3

---

Title: Coresets from Trajectories: Selecting Data via Correlation of Loss Differences

Abstract: Deep learning models achieve state-of-the-art performance across domains but face scalability challenges in real-time or resource-constrained scenarios. To address this, we propose Correlation of Loss Differences ($\mathtt{CLD}$), a simple and scalable metric for coreset selection that identifies the most impactful training samples by measuring their alignment with the loss trajectories of a held-out validation set. $\mathtt{CLD}$ is highly efficient, requiring only per-sample loss values computed at training checkpoints, and avoiding the costly gradient and curvature computations used in many existing subset selection methods. We develop a general theoretical framework that establishes convergence guarantees for $\mathtt{CLD}$-based coresets, demonstrating that the convergence error is upper-bounded by the alignment of the selected samples and the representativeness of the validation set. On CIFAR-100 and ImageNet-1k, $\mathtt{CLD}$-based coresets typically outperform or closely match state-of-the-art methods across subset sizes, and remain within 1\% of more computationally expensive baselines even when not leading. $\mathtt{CLD}$ transfers effectively across architectures (ResNet, VGG, DenseNet), enabling proxy-to-target selection with $<1\%$ degradation. Moreover, $\mathtt{CLD}$ is stable when using only early checkpoints, incurring negligible accuracy loss. Finally, $\mathtt{CLD}$ exhibits inherent bias reduction via per-class validation alignment, obviating the need for additional stratified sampling. Together, these properties make $\mathtt{CLD}$ a principled, efficient, stable, and transferable tool for scalable dataset optimization.

URL: https://openreview.net/forum?id=QY0pbZTWJ9

---

Title: RT2I-Bench: Evaluating Robustness of Text-to-Image Systems Against Adversarial Attacks

Abstract: Text-to-Image (T2I) systems have demonstrated impressive abilities in the generation of images from text descriptions. However, these systems remain susceptible to adversarial prompts—carefully crafted input manipulations that can result in misaligned or even toxic outputs. This vulnerability highlights the need for systematic evaluation and development of attack strategies that exploit these weaknesses, as well as defense mechanisms that safeguard T2I models. This work introduces RT2I-Bench, a comprehensive benchmark designed to assess the robustness of T2I systems against adversarial attacks. The benchmark serves two primary purposes. First, it provides a structured evaluation of various adversarial attacks, examining their effectiveness, transferability, stealthiness and potential for generating misaligned or toxic outputs, as well as assessing the resilience of state-of-the-art T2I models to such attacks. We observe that state-of-the-art T2I systems are vulnerable to adversarial prompts, with the most effective attacks achieving success rates of over 60\% across the majority of T2I models we tested. Second, RT2I-Bench enables the creation of a set of strong adversarial prompts (consisting of 1,439 that induce misaligned or targeted outputs and 173 that induce toxic outputs), which are effective across a wide range of systems. This dataset offers a valuable resource for robustness testing and defense evaluation. Finally, our benchmark is designed to be extensible, enabling the seamless addition of new attack techniques, T2I models, and evaluation metrics. This flexible framework provides an automated and scalable solution for robustness assessment and adversarial prompt generation in T2I systems.

URL: https://openreview.net/forum?id=ZUiWjEouSf

---

Title: DREAMS: Preserving both Local and Global Structure in Dimensionality Reduction

Abstract: Dimensionality reduction techniques are widely used for visualizing high-dimensional data in two dimensions. Existing methods are typically designed to preserve either local (e.g. $t$-SNE, UMAP) or global (e.g. MDS, PCA) structure of the data, but none of the established methods can represent both aspects well. In this paper, we present DREAMS (Dimensionality Reduction Enhanced Across Multiple Scales), a method that combines the local structure preservation of $t$-SNE with the global structure preservation of PCA via a simple regularization term. Our approach generates a spectrum of embeddings between the locally well-structured $t$-SNE embedding and the globally well-structured PCA embedding, efficiently balancing both local and global structure preservation. We benchmark DREAMS across seven real-world datasets, including five from single-cell transcriptomics and one from population genetics, showcasing qualitatively and quantitatively its superior ability to preserve structure across multiple scales compared to previous approaches.

URL: https://openreview.net/forum?id=xpGu3Sichc

---

Title: Node Embeddings via Neighbor Embeddings

Abstract: Node embeddings are a paradigm in non-parametric graph representation learning, where graph nodes are embedded into a given vector space to enable downstream processing. State-of-the-art node-embedding algorithms, such as DeepWalk and node2vec, are based on random-walk notions of node similarity and on contrastive learning. In this work, we introduce the graph neighbor-embedding (graph NE) framework that directly pulls together embedding vectors of adjacent nodes without relying on any random walks. We show that graph NE strongly outperforms state-of-the-art node-embedding algorithms in terms of local structure preservation. Furthermore, we apply graph NE to the 2D node-embedding problem, obtaining graph layouts that also outperform existing graph-layout algorithms.

URL: https://openreview.net/forum?id=8APIU9cauZ

---

Title: Pre-trained Language Models Improve the Few-shot Prompt Ability of Decision Transformer

Abstract: Decision Transformer (DT) has emerged as a promising class of algorithms in offline reinforcement learning (RL) tasks, leveraging pre-collected datasets and Transformer's capability to model long sequences. Recent works have demonstrated that using parts of trajectories from training tasks as prompts in DT enhances its performance on unseen tasks, giving rise to Prompt-DT methods. However, collecting data from specific environments can be both costly and unsafe in many scenarios, leading to suboptimal performance and limited few-shot prompt abilities due to the data-hungry nature of Transformer-based models. Additionally, the limited datasets used in pre-training make it challenging for Prompt-DT type of methods to distinguish between various RL tasks through prompts alone. To address these challenges, we introduce the Language model-initialized Prompt Decision Transformer (LPDT) framework, which leverages pretrained language models providing rich prior knowledge for RL tasks and fine-tunes the sequence model using Low-rank Adaptation (LoRA) for meta-RL problems. We further incorporate prompt regularization to effectively differentiate between tasks based on prompt feature representations. Comprehensive empirical studies demonstrate that initializing with a pre-trained language model provides the prior knowledge and achieves a similar performance with Prompt-DT under only $10\%$ data. We also provide a thorough ablation study to validate the effectiveness of each component, including sequence modeling, language models, prompt regularizations, and prompt strategies.

URL: https://openreview.net/forum?id=k520i3XEMK

---

Title: Improving Detection of Watermarked Language Models

Abstract: Watermarking has recently emerged as an effective strategy for detecting the generations of large language models (LLMs). The strength of a watermark typically depends strongly on the entropy afforded by the language model and the set of input prompts. However, entropy can be quite limited in practice, especially for models that are post-trained, for example via instruction tuning or reinforcement learning from human feedback (RLHF), which makes detection based on watermarking alone challenging. In this work, we investigate whether detection can be improved by combining watermark detectors with \emph{non-watermark} ones. We explore a number of \emph{hybrid} schemes that combine the two, observing performance gains over either class of detector under a wide range of experimental conditions.

URL: https://openreview.net/forum?id=6nmztBgngB

---

Title: Designing a Conditional Prior Distribution for Flow-Based Generative Models

Abstract: Flow-based generative models have recently shown impressive performance for conditional generation tasks, such as text-to-image generation. However, current methods transform a general unimodal noise distribution to a specific mode of the target data distribution. As such, every point in the initial source distribution can be mapped to every point in the target distribution, resulting in long average paths. To this end, in this work, we tap into a non-utilized property of conditional flow-based models: the ability to design a non-trivial prior distribution. Given an input condition, such as a text prompt, we first map it to a point lying in data space, representing an “average" data point with the minimal average distance to all data points of the same conditional mode (e.g., class). We then utilize the flow matching formulation to map samples from a parametric distribution centered around this point to the conditional target distribution. Experimentally, our method significantly improves training times and generation efficiency (FID, KID and CLIP alignment scores) compared to baselines, producing high quality samples using fewer sampling steps.

URL: https://openreview.net/forum?id=Teh9Bq4giF

---

Title: Towards shutdownable agents via stochastic choice

Abstract: The POST-Agents Proposal (PAP) is an idea for ensuring that advanced artificial agents never resist shutdown. A key part of the PAP is using a novel ‘Discounted Reward for Same-Length Trajectories (DReST)’ reward function to train agents to (1) pursue goals effectively conditional on each trajectory-length (be 'USEFUL'), and (2) choose stochastically between different trajectory-lengths (be NEUTRAL' about trajectory-lengths). In this paper, we propose evaluation metrics for USEFULNESS and NEUTRALITY. We use a DReST reward function to train simple agents to navigate gridworlds, and we find that these agents learn to be USEFUL and NEUTRAL. Our results thus provide some initial evidence that DReST reward functions could train advanced agents to be USEFUL and NEUTRAL. Our theoretical work suggests that these agents would be useful and shutdownable.

URL: https://openreview.net/forum?id=j5Qv7KdWBn

---

Title: Bi-level Hierarchical Neural Contextual Bandits for Online Recommendation

Abstract: Contextual bandit algorithms aim to identify the optimal choice among a set of candidate arms, based on their contextual information. Among others, neural contextual bandit algorithms have demonstrated generally superior performance compared to conventional linear and kernel-based methods. Nevertheless, neural methods can be inherently unsuitable for handling a large number of candidate arms due to their high computational cost when performing principled exploration. Motivated by the widespread availability of arm category information (e.g., movie genres, retailer types), we formulate contextual bandits as a bi-level online recommendation problem, and propose a novel neural bandit framework, named $\text{H}_{2}\text{N-Bandit}$, which utilizes a bi-level hierarchical neural architecture to mitigate the substantial computational cost found in conventional neural bandit methods. To demonstrate its theoretical effectiveness, we provide regret analysis under general over-parameterization settings, along with a guarantee for category-level recommendation. To illustrate its effectiveness and efficiency, we conduct extensive experiments on multiple real-world data sets, highlighting that $\text{H}_{2}\text{N-Bandit}$ can significantly reduce the computational cost over existing strong non-linear baselines, while achieving better or comparable performance under online recommendation settings.

URL: https://openreview.net/forum?id=k3XsA75SGv

---

Title: A Modular Abstraction for Integrating Domain Rules into Deep Learning Models

Abstract: Domain-specific knowledge can often be expressed as suggestive rules defined over subgroups of data. Such rules, when encoded as hard constraints, are often not directly compatible with deep learning frameworks that train neural networks over batches of data. Also, domain-experts often use heuristics that should not be encoded as logical rules. In this work, we propose a framework to capture domain-experts' knowledge as domain-specific rules over subgroups of data, and to leverage such rules in training deep learning models using the modular components of regularization, data augmentation, and parameter optimization. This translation of domain knowledge into custom primitives that can be augmented to existing state-of-the-art deep learning models improves the ability of domain experts to interpret and express model behavior, intervene through changes in the modeling specifications, and improve the overall performance of the model as compared to existing frameworks that incorporate deterministic declarative predicates. On one synthetic and three real-world tasks, we show that our method allows iterative refinement and is demonstrably more accurate.

URL: https://openreview.net/forum?id=KicRPZsIDH

---

Title: Unreasonable effectiveness of LLM reasoning: a doubly cautionary tale of temporal question-answering

Abstract: The remarkable success of Large Language Models in modeling both the syntax and the semantics of language has prompted a body of research into language-adjacent abilities, most notably commonsense reasoning.
As LLMs' performance continues to advance on successive benchmarks, we turn to temporal reasoning, which lags somewhat behind other tasks due to its more complex logic.
We start from previous work, where authors successfully induce (apparent) reasoning by breaking down the problem into a two-step procedure of temporal graph extraction and subsequent reasoning.
Specifically, in the first step an LLM is prompted to parse a natural language description into a semi-structured timeline of events; and in the second step, it is given the extracted timeline and prompted to answer a temporal reasoning question.
We conjecture that this procedure presents two separate opportunities for introducing errors and further hypothesise that a Neuro-symbolic approach should help in this matter.
We follow the recent trend of using external executors in concert with LLMs to carry out exact reasoning and verification.
We see the reasoning step of the original two-step procedure as a natural target for a symbolic solver and design a rule-based solution for Temporal Question-Answering, drawing on ideas from Allen’s Interval Algebra.
To our surprise, we find that our rule-based reasoner does not improve beyond the previously reported, purely neural solution.
It appears that both our approach and the previous method operate at around the limits of achievable performance, imposed by the correctness of information extraction.
Such a result seems to suggest that a non-symbolic LLM is capable of symbolic-level reasoning, although upon further investigation we discover that not to be the case.
It is not that the neural solution makes no reasoning mistakes, but rather that the LLM manages to compensate for some of its erroneous replies by `short-cutting' to the correct answer in other questions; a.k.a. not reasoning but guessing.
Although the effect is not pronounced performance-wise, we feel it is conceptually important: as we argue, production of correct answers is not a measure of reasoning.

URL: https://openreview.net/forum?id=1DkD0Nd8Rd

---

Title: Real-Time Deepfake Detection in the Real World

Abstract: Recent improvements in generative AI made synthesizing fake images easy; as they can be used to cause harm, it is crucial to develop accurate techniques to identify them. This paper introduces "Locally Aware Deepfake Detection Algorithm" (LaDeDa), that accepts a single 9 x9 image patch and outputs its deepfake score. The image deepfake score is the pooled score of its patches. With merely patch-level information, LaDeDa significantly improves over the state-of-the-art, achieving around 99% mAP on current benchmarks. Owing to the patch-level structure of LaDeDa, we hypothesize that the generation artifacts can be detected by a simple model. We therefore distill LaDeDa into Tiny-LaDeDa, a highly efficient model consisting of only 4 convolutional layers. Remarkably, Tiny-LaDeDa has 375x fewer FLOPs and is 10,000x more parameter-efficient than LaDeDa, allowing it to run efficiently on edge devices with a minor decrease in accuracy. These almost-perfect scores raise the question: is the task of deepfake detection close to being solved? Perhaps surprisingly, our investigation reveals that current training protocols prevent methods from generalizing to real-world deepfakes extracted from social media. To address this issue, we introduce WildRF, a new deepfake detection dataset curated from several popular social networks. Our method achieves the top performance of 93.7% mAP on WildRF, however the large gap from perfect accuracy shows that reliable real-world deepfake detection is still unsolved.

URL: https://openreview.net/forum?id=ibmmuUJTCx

---

Title: Prompt Engineering Techniques for Language Model Reasoning Lack Replicability

Abstract: As large language models (LLMs) are integrated into everyday applications, research into prompt engineering techniques (PET) to improve these models’ behavior has surged. How- ever, clear methodological guidelines for evaluating these techniques are lacking. This raises concerns about the replicability and generalizability of the prompt engineering techniques’ benefits. We support our concerns with a series of replication experiments focused on zero- shot prompt engineering techniques purported to influence reasoning abilities in LLMs. We tested GPT-3.5, GPT-4o, Gemini 1.5 Pro, Claude 3 Opus, Llama 3, Vicuna, and BLOOM on the chain-of-thought, EmotionPrompting, Sandbagging, Re-Reading, Rephrase- and-Respond (RaR), and ExpertPrompting prompt engineering techniques. We applied them on manually double-checked subsets of reasoning benchmarks including Common- senseQA, CRT, NumGLUE, ScienceQA, and StrategyQA. Our findings reveal a general lack of statistically significant differences across nearly all techniques tested, highlighting, among others, several methodological weaknesses in previous research. To counter these issues, we propose recommendations for establishing sound benchmarks, and designing rigorous exper- imental frameworks to ensure accurate and reliable assessments of model outputs.

URL: https://openreview.net/forum?id=bgjR5bM44u

---

Title: Clus-UCB: A Near-Optimal Algorithm for Clustered Bandits

Abstract: We study a stochastic multi-armed bandit setting where arms are partitioned into known clusters, such that the mean rewards of arms within a cluster differ by at most a known threshold. While the clustering structure is known a priori, the arm means are unknown. We derive an asymptotic lower bound on the regret that improves upon the classical bound of Lai & Robbins (1985). We then propose Clus-UCB, an efficient algorithm that closely matches this lower bound asymptotically. Clus-UCB is designed to exploit the clustering structure and introduces a new index to evaluate an arm, which depends on other arms within the cluster. In this way, arms share information among each other. We present simulation results of our algorithm and compare its performance against KL-UCB and other well known algorithms for bandits with dependent arms. Finally, we address some limitations of this work and conclude by mentioning some possible future research.

URL: https://openreview.net/forum?id=QDMvPO9WJT

---

Title: Model Alignment Search

Abstract: When can we say that two neural systems are the same? What nuances do we miss when we fail to causally probe the representations of the systems? In this work, we introduce a method for connecting neural representational similarity to behavior through causal interventions. The method learns transformations that find an aligned subspace in which behavioral information can be interchanged between multiple distributed networks' representations. We first show that the method can be used to transfer the behavior from one frozen Neural Network (NN) to another in a manner similar to model stitching, and we show how the method can differ from correlative similarity measures like Representational Similarity Analysis. Next, we empirically and theoretically show how the method can be equivalent to model stitching when desired, or it can take a form that has a more restrictive focus to shared causal information; in both forms, it reduces the number of required matrices for a comparison of n models to be linear in n. We then present a case study on number-related tasks showing that the method can be used to examine specific subtypes of causal information, and we present another case study showing that the method can reveal toxicity in fine-tuned DeepSeek-r1-Qwen-1.5B models. Lastly, we show how to augment the loss with a counterfactual latent auxiliary objective to improve causal relevance when one of the two networks is causally inaccessible (as is often the case in comparisons with biological networks). We use our results to encourage the use of causal methods in neural similarity analyses and to suggest future explorations of network similarity methodology for model misalignment.

URL: https://openreview.net/forum?id=I9shNCSmCU

---

Title: AC$\oplus$DC search: behind the winning solution to the Flywire graph-matching challenge

Abstract: This paper describes the alternating continuous and discrete combinatorial (AC$\oplus$DC) optimizations behind the winning solution to the Flywire Ventral Nerve Cord Matching Challenge. The challenge was organized by the Princeton Neuroscience Institute and held over three months, ending on January 31, 2025. During this period, the challenge attracted teams of researchers with expertise in machine learning, high-performance computing, graph data mining, biological network analysis, and quadratic assignment problems. The goal of the challenge was to align the connectomes of a male and female fruit fly, and more specifically, to determine a one-to-one correspondence between the neurons in their ventral nerve cords. The connectomes were represented as large weighted graphs, and the challenge was posed as a problem in graph matching: how does one find a permutation that maps the nodes of one graph onto the nodes of another? The winning solution to the challenge alternated between two complementary approaches to graph matching—the first, a combinatorial optimization over the symmetric group of permutations, and the second, a continuous relaxation of this problem to the space of doubly stochastic matrices. For the latter, the doubly stochastic matrices were optimized by combining Frank-Wolfe methods with a fast preconditioner to solve the linear assignment problem at each iteration. We provide a complete implementation of these methods with a few hundred lines of code in MATLAB. Notably, this implementation obtains a winning score to the challenge in less than 15 minutes on a laptop computer.

URL: https://openreview.net/forum?id=8MjCOMyaDf

---

Title: Transformers as Implicit State Estimators: In-Context Learning in Dynamical Systems

Abstract: Predicting the behavior of a dynamical system from noisy observations of its past outputs is a classical problem encountered across engineering and science. For linear systems with Gaussian inputs, the Kalman filter -- the best linear minimum mean-square error estimator of the state trajectory -- is optimal in the Bayesian sense. For nonlinear systems, Bayesian filtering is typically approached using suboptimal heuristics such as the Extended Kalman Filter (EKF), or numerical methods such as particle filtering (PF). In this work, we show that transformers, employed in an in-context learning (ICL) setting, can implicitly infer hidden states in order to predict the outputs of a wide family of dynamical systems, without test-time gradient updates or explicit knowledge of the system model. Specifically, when provided with a short context of past input–output pairs and, optionally, system parameters, a frozen transformer accurately predicts the current output. In linear-Gaussian regimes, its predictions closely match those of the Kalman filter; in nonlinear regimes, its performance approaches that of EKF and PF. Moreover, prediction accuracy degrades gracefully when key parameters, such as the state-transition matrix, are withheld from the context, demonstrating robustness and implicit parameter inference. These findings suggest that transformer in-context learning provides a flexible, non-parametric alternative for output prediction in dynamical systems, grounded in implicit latent-state estimation.

URL: https://openreview.net/forum?id=hIMK5MvGkP

---

Title: StructEval: Benchmarking LLMs' Capabilities to Generate Structural Outputs

Abstract: As Large Language Models (LLMs) become integral to software development workflows, their ability to generate structured outputs has become critically important. We introduce $\textbf{StructEval}$, a comprehensive benchmark for evaluating LLMs' capabilities in producing both non-renderable (JSON, YAML, CSV) and renderable (HTML, React, SVG) structured formats. Unlike prior benchmarks, StructEval systematically evaluates structural fidelity across diverse formats through two paradigms: $\textbf{1)}$ generation tasks, producing structured output from natural language prompts, and $\textbf{2)}$ conversion tasks, translating between structured formats. Our benchmark encompasses 18 formats and 44 types of task, with novel metrics for format adherence and structural correctness. Results reveal significant performance gaps—even state-of-the-art models like o1-mini achieve only $75.58$ average score, with open-source alternatives lagging approximately $10$ points behind. We find generation tasks more challenging than conversion tasks, and producing correct visual content more difficult than generating text-only structures.

URL: https://openreview.net/forum?id=buDwV7LUA7

---

Title: The Geometry of Stability: A Cohomological View on Preference Cycles and Algorithmic Robustness

Abstract: Algorithmic stability—the robustness of predictions to training data perturbations—is fundamental to reliable machine learning. While methods like bagging, regularization, and inflated operators improve stability, they appear as disconnected techniques. We propose a unified mathematical framework demonstrating that algorithmic instability often arises from fundamental inconsistencies in local data preferences, mathematically analogous to Condorcet cycles in social choice theory. We formalize these inconsistencies as cohomological obstructions ($H^1 \neq 0$), leveraging established connections between social choice theory and algebraic topology. This framework reveals bagging as a strategy for obstruction prevention (smoothing the preference landscape) and inflated operators as a strategy for obstruction resolution (target space enlargement). Furthermore, we derive a novel technique from this framework, obstruction-aware regularization, which directly enforces mathematical consistency. We provide direct empirical validation for our claims. First, we demonstrate that engineered Condorcet cycles induce high instability in standard methods, which is resolved by inflated operators. Second, using Hodge decomposition, we confirm that bagging significantly reduces the magnitude of cohomological obstructions. Third, we show that our proposed obstruction-aware regularization successfully reduces mathematical inconsistencies and yields substantial improvements across multiple metrics of algorithmic stability.

URL: https://openreview.net/forum?id=rFqsgVXZYO

---

Title: RIZE: Regularized Imitation Learning via Distributional Reinforcement Learning

Abstract: We propose a novel Inverse Reinforcement Learning (IRL) method that mitigates the rigidity of fixed reward structures and the limited flexibility of implicit reward regularization. Building on the Maximum Entropy IRL framework, our approach incorporates a squared temporal-difference (TD) regularizer with adaptive targets that evolve dynamically during training, thereby imposing adaptive bounds on recovered rewards and promoting robust decision-making. To capture richer return information, we integrate distributional RL into the learning process. Empirically, our method achieves expert-level performance on complex MuJoCo tasks, surpassing baseline methods on the Humanoid task with 3 demonstrations. Extensive experiments and ablation studies further validate the effectiveness of the approach and provide insights into reward dynamics in imitation learning.

URL: https://openreview.net/forum?id=a6DWqXJZCZ

---

Title: ADMIRE-BayesOpt: Accelerated Data MIxture RE-weighting for Language Models with Bayesian Optimization

Abstract: Determining the optimal data mixture for large language model training remains a challenging problem with an outsized impact on performance. In practice, language model developers continue to rely on heuristic exploration since no learning-based approach has emerged as a reliable solution. In this work, we propose to view the selection of training data mixtures as a black-box hyperparameter optimization problem, for which Bayesian Optimization is a well-established class of appropriate algorithms. Firstly, we cast data mixture learning as a sequential decision-making problem, in which we aim to find a suitable trade-off between the computational cost of training exploratory (proxy-) models and final mixture performance. Secondly, we systematically explore the properties of transferring mixtures learned at a small scale to larger-scale experiments, providing insights and highlighting opportunities for research at a modest scale. By proposing Multi-fidelity Bayesian Optimization as a suitable method in this common scenario, we introduce a natural framework to balance experiment cost with model fit, avoiding the risks of overfitting to smaller scales while minimizing the number of experiments at high cost. We present results for pre-training and instruction finetuning across models ranging from 1 million to 7 billion parameters, varying from simple architectures to state-of-the-art models and benchmarks spanning dozens of datasets. We demonstrate consistently strong results relative to a wide range of benchmarks, showing a speed-ups of over 500% in determining the best data mixture on our largest experiments relative to recent baselines. In addition, we broaden access to research by sharing ADMIRE IFT Runs, a dataset of 460 full training & evaluation runs reproducible post-training pipelines worth over 13,000 GPU hours, greatly reducing the cost of conducting research in this area. Finally, we highlight rich opportunities for future research in this area, helping bridge the gap towards a comprehensive understanding of the broader effects of training data on model generalization.

URL: https://openreview.net/forum?id=0Euvm9zDpu

---

Title: Improved DDIM Sampling with Moment Matching Gaussian Mixtures

Abstract: We propose using a Gaussian Mixture Model (GMM) as reverse transition operator (kernel) within the Denoising Diffusion Implicit Models (DDIM) framework, which is one of the most widely used approaches for accelerated sampling from pre-trained Denoising Diffusion Probabilistic Models (DDPM). Specifically we match the first and second order central moments of the DDPM forward marginals by constraining the parameters of the GMM. We see that moment matching is sufficient to obtain samples with equal or better quality than the original DDIM with Gaussian kernels. We provide experimental results with unconditional models trained on CelebAHQ and FFHQ, class-conditional models trained on ImageNet, and text-to-image generation using Stable Diffusion v2.1 on COYO700M datasets respectively. Our results suggest that using the GMM kernel leads to significant improvements in the quality of the generated samples when the number of sampling steps is small, as measured by FID and IS metrics. For example on ImageNet 256x256, using 10 sampling steps, we achieve a FID of 6.94 and IS of 207.85 with a GMM kernel compared to 10.15 and 196.73 respectively with a Gaussian kernel. Further, we derive novel SDE samplers for rectified flow matching models and experiment with the proposed approach. We see improvements using both 1-rectified flow and 2-rectified flow models.

URL: https://openreview.net/forum?id=CdSPjfmrQN

---

Title: Fast weight programming and linear transformers: from machine learning to neurobiology

Abstract: Recent advances in artificial neural networks for machine learning, and language modeling in particular, have established a family of recurrent neural network (RNN) architectures that, unlike conventional RNNs with vector-form hidden states, use two-dimensional (2D) matrix-form hidden states. Such 2D-state RNNs, known as Fast Weight Programmers (FWPs), can be interpreted as a neural network whose synaptic weights (called fast weights) dynamically change over time as a function of input observations, and serve as short-term memory storage; corresponding synaptic weight modifications are controlled or programmed by another network (the programmer) whose parameters are trained (e.g., by gradient descent). In this Primer, we review the technical foundations of FWPs, their computational characteristics, and their connections to transformers and state space models. We also discuss connections between FWPs and models of synaptic plasticity in the brain, suggesting a convergence of natural and artificial intelligence.

URL: https://openreview.net/forum?id=TDG8EkNmQR

---

Title: Symbolic Quantile Regression for the Interpretable Prediction of Conditional Quantiles

Abstract: Symbolic Regression (SR) is a well-established framework for generating interpretable or white-box predictive models.
Although SR has been successfully applied to create interpretable estimates of the average of the outcome, it is currently not well understood how it can be used to estimate the relationship between variables at other points in the distribution of the target variable. Such estimates of e.g. the median or an extreme value provide a fuller picture of how predictive variables affect the outcome and are necessary in high-stakes, safety-critical application domains. This study introduces Symbolic Quantile Regression (SQR), an approach to predict conditional quantiles with SR. In an extensive evaluation, we find that SQR outperforms transparent models and performs comparably to a strong black-box baseline without compromising transparency. We also show how SQR can be used to explain differences in the target distribution by comparing models that predict extreme and central outcomes in an airline fuel usage case study. We conclude that SQR is suitable for predicting conditional quantiles and understanding interesting feature influences at varying quantiles.

URL: https://openreview.net/forum?id=x9OYbyPJOG

---

Title: Incentivizing High-quality Participation From Federated Learning Agents

Abstract: Federated learning (FL) provides a promising paradigm for facilitating collaboration between multiple clients that jointly learn a global model without directly sharing their local data. However, existing research suffers from two caveats: 1) From the perspective of agents, voluntary and unselfish participation is often assumed. But self-interested agents may opt out of the system or provide low-quality contributions without proper incentives; 2) From the mechanism designer's perspective, the aggregated models can be unsatisfactory as the existing game-theoretical federated learning approach for data collection ignores the potential heterogeneous effort caused by contributed data.
To alleviate above challenges, we propose an incentive-aware framework for agent participation that considers data heterogeneity to accelerate the convergence process. Specifically, we first introduce the notion of Wasserstein distance to explicitly illustrate the heterogeneous effort and reformulate the existing upper bound of convergence. To induce truthful reporting from agents, we analyze and measure the generalization error gap of any two agents by leveraging the peer prediction mechanism to develop score functions. We further present a two-stage Stackelberg game model that formalizes the process and examines the existence of equilibrium. Extensive experiments on real-world datasets demonstrate the effectiveness of our proposed mechanism.

URL: https://openreview.net/forum?id=PeaEnCWAQa

---

Title: Unbiased Stochastic Optimization for Gaussian Processes on Finite Dimensional RKHS

Abstract: Current methods for stochastic hyperparameter learning in Gaussian Processes (GPs) rely
on approximations, such as computing biased stochastic gradients or using inducing points in
stochastic variational inference. However, when using such methods we are not guaranteed
to converge to a stationary point of the true marginal likelihood. In this work, we propose
algorithms for exact stochastic inference of GPs with kernels that induce a Reproducing
Kernel Hilbert Space (RKHS) of moderate finite dimension. Our approach can also be
extended to infinite dimensional RKHSs at the cost of forgoing exactness. Both for finite and
infinite dimensional RKHSs, our method achieves better experimental results than existing
methods when memory resources limit the feasible batch size and the possible number of
inducing points.

URL: https://openreview.net/forum?id=nVRpd28Fms

---

Title: The impact of imbalanced datasets on Deep Neural Network predictions: A case study in scramjet performance

Abstract: Robust aerodynamic predictions for hypersonic vehicles increasingly rely on existing deep‑learning tools. However, imbalanced datasets — often resulting from limited experimental data or insufficient coverage of operational conditions — can compromise model reliability and introduce bias into predictions. This work offers an application‑centered account of how a feed‑forward multilayer perceptron (PyTorch implementation) behaves when trained on (i) a data‑rich yet operationally imbalanced set of scramjet simulations and (ii) a deliberately balanced counterpart generated with a conventional metaheuristic (MH) sampling scheme, but with a lower sample count. Without altering network architecture, loss function, or optimizer, we expose a clear trade‑off: the imbalanced model achieves a 14% lower root mean square error (RMSE) but produces thrust predictions that violate first‑principles trends, whereas the balanced model sacrifices a small amount of numerical accuracy to maintain physical coherence across Mach–altitude space. These results illuminate both the strength (high statistical accuracy) and the weakness (loss of physical fidelity under bias) of off‑the‑shelf deep neural networks (DNNs) when data coverage is uneven. The findings serve as a cautionary example for practitioners who might otherwise deploy such models uncritically, and underscore the methodological importance of rigorous dataset diagnostics — rather than chasing novel algorithms — for reliable AI adoption in aerospace design.

URL: https://openreview.net/forum?id=0kHvatNcGB

---

Title: Model Debiasing by Learnable Data Augmentation

Abstract: Deep Neural Networks are well known for efficiently fitting training data, yet experiencing poor generalization capabilities whenever some kind of bias dominates over the actual task labels, resulting in models learning “shortcuts”. In essence, such models are often prone to learn spurious correlations between data and labels. In this work, we tackle the problem of learning from biased data in the very realistic unsupervised scenario, i.e., when the bias is unknown. This is a much harder task as compared to the supervised case, where auxiliary, bias-related annotations, can be exploited in the learning process. This paper proposes a novel 2-stage learning pipeline featuring a data augmentation strategy able to regularize the training. First, biased/unbiased samples are identified by training over-biased models.
Second, such subdivision (typically noisy) is exploited within a data augmentation framework, properly combining the original samples while learning mixing parameters, which has a regularization effect. Experiments on synthetic and realistic biased datasets show state-of-the-art classification accuracy, outperforming competing methods, ultimately proving robust performance on both biased and unbiased examples. Notably, being our training method totally agnostic to the level of bias, it also positively affects performance for any, even apparently unbiased, dataset, thus improving the model generalization regardless of the level of bias (or its absence) in the data.

URL: https://openreview.net/forum?id=3ac7heNftC

---

Title: Cropping outperforms dropout as an augmentation strategy for training self-supervised text embeddings

Abstract: Text embeddings, i.e. vector representations of entire texts, play an important role in many NLP applications, such as retrieval-augmented generation, sentiment analysis, clustering, or visualizing collections of texts for data exploration. Currently, top-performing embedding models are derived from pre-trained language models via extensive supervised fine-tuning using curated text pairs. This contrasts with computer vision, where self-supervised training based on data augmentations has demonstrated remarkable success. Here we systematically compare the two most well-known augmentation strategies for positive pair generation in contrastive learning of text embeddings. We assess embedding quality on MTEB and additional in-domain evaluations and show that cropping augmentation strongly outperforms the dropout-based approach. We find that on out-of-domain data, the quality of resulting embeddings is below the supervised SOTA models, but for in-domain data, self-supervised fine-tuning produces high-quality text embeddings after very short fine-tuning, sometimes only marginally below the supervised SOTA. Finally, we show that representation quality increases towards the last transformer layers, which undergo the largest change during fine-tuning; and that fine-tuning only those last layers is sufficient to reach similar embedding quality.

URL: https://openreview.net/forum?id=gVRsIh9x7W

---

Title: TabRep: Training Tabular Diffusion Models with a Simple and Effective Continuous Representation

Abstract: Diffusion models have been the predominant generative model for tabular data generation. However, they face the conundrum of modeling under a separate versus a unified data representation. The former encounters the challenge of jointly modeling all multi-modal distributions of tabular data in one model. While the latter alleviates this by learning a single representation for all features, it currently leverages sparse suboptimal encoding heuristics and necessitates additional computation costs. In this work, we address the latter by presenting TabRep, a tabular diffusion architecture trained with a unified continuous representation. To motivate the design of our representation, we provide geometric insights into how the data manifold affects diffusion models. The key attributes of our representation are composed of its density, flexibility to provide ample separability for nominal features, and ability to preserve intrinsic relationships. Ultimately, TabRep provides a simple yet effective approach for training tabular diffusion models under a continuous data manifold. Our results showcase that TabRep achieves superior performance across a broad suite of evaluations. It is the first to synthesize tabular data that exceeds the downstream quality of the original datasets while preserving privacy and remaining computationally efficient.

URL: https://openreview.net/forum?id=yRbtFEh2OP

---