Get Started
Home
Topics
Search
Library
Research questionHow can we tell whether LLMs follow coherent, human-like prerequisite relationships in mathematical reasoning?A model can answer individual mathematics questions correctly while violating expected prerequisite dependencies or failing to use related knowledge in context. These structural inconsistencies may remain hidden by accuracy-based and LLM-as-judge evaluations.
AI
Evaluation & Benchmarks
Reasoning
Research Paper
Latest papersRecent research connected to this question, newest first.Do LLMs Exhibit Coherent Knowledge Structures in Mathematical Reasoning? A Perspective from Knowledge Space TheoryThe evidence concerns a Knowledge Space Theory–grounded behavioral analysis of eight open- and closed-source LLMs on mathematical reasoning, compared with real human learners. It examines dependency violations, use of related contextual knowledge on dependent questions, and overlap among models’ inferred knowledge distributions; the findings provide behavioral evidence rather than direct evidence about internal representations.research paper · Sep 4, 2026
Related questions
How can mathematical-reasoning LLMs learn to construct counterexamples that reveal conceptual understanding?How can we tell whether LLM hidden-state geometry reflects reasoning operations rather than lexical or positional cues?How can LLMs interleave reasoning with reliable step-level self-critique without a separate verifier?What determines whether an LLM transfers arithmetic reasoning across numeric and verbal prompts?