Predicate Label Inconsistencies - LLMs4OL 2026 Reuse Task

[Upal Bhattacharya]

Dear LLMs4OL 2026 Organizing Committee,

We are the HybridAI-Squared team participating in the Reuse Task. We observed inconsistencies in the name of predicates between the released test and training data and we were hoping to get clarifications on this.

Certain predicates in the initial and extended triples e.g. 'is defined by' are whitespace-separated in the training data (train_task_b.json) but are hyphenated i.e. 'is-defined-by' in the test data (test_task_b_input.json). In the submissions, is there a predicate label convention that should be followed or can either convention be used?

Thank you for organizing this challenge.

Kind regards,

HybridAI-Squared team

[LLMs4OL Challenge]

Dear HybridAI-Squared team,

Thank you for your question and for carefully reviewing the datasets.

The observed difference in predicate formatting (e.g., `is defined by` versus `is-defined-by`) is not expected to affect the objective of the Reuse Task. The task is designed to encourage systems to reuse the provided primitive ontology and its context to induce the missing triples based on the provided artifacts, rather than focusing on minor formatting variations in predicate labels. Additionally, our evaluation accounts for these formatting differences, so using hyphenated or whitespace-separated predicate labels will not negatively impact the evaluation. Our recommendation is to follow the primitive ontology and predicate conventions provided in the input ontology that your system is extending. In other words, your model should learn and preserve the predicate style of the ontology it is working with, rather than attempting to normalize predicate labels across the datasets.

We hope this clarifies the intended behavior. Please feel free to reach out if you have any further questions.

Kind regards,

Hamed

--
You received this message because you are subscribed to the Google Groups "LLMs4OL Challenge" group.
To unsubscribe from this group and stop receiving emails from it, send an email to llms4ol-challe...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/llms4ol-challenge/a168c76a-ef7b-412d-afe4-3ae2a0618a1an%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.