Fundamental Limitations of Large Language Models in Reasoning
60 views
Skip to first unread message
Pei Wang
Apr 2, 2025, 8:21:42 AMApr 2
to open-nars
TALK: Fundamental Limitations of Large Language Models in Reasoning
ABSTRACT: This talk analyzes models of reasoning. By comparing two interpretations of "inference rule" (derivation vs. implication), it can be shown that though Large Language Models solve many problems that usually demand human reasoning capability, the mechanism used is fundamentally different from what is considered as "reasoning" in the related disciplines. This difference has important theoretical and practical consequences.
VIDEO: https://www.youtube.com/watch?v=PXqnqmm_UfI
PPT: https://cis.temple.edu/tagit/presentations/ReasoningModels.pdf
ABSTRACT: This talk analyzes models of reasoning. By comparing two interpretations of "inference rule" (derivation vs. implication), it can be shown that though Large Language Models solve many problems that usually demand human reasoning capability, the mechanism used is fundamentally different from what is considered as "reasoning" in the related disciplines. This difference has important theoretical and practical consequences.
VIDEO: https://www.youtube.com/watch?v=PXqnqmm_UfI
PPT: https://cis.temple.edu/tagit/presentations/ReasoningModels.pdf
Reply all
Reply to author
Forward
0 new messages