A Four-Stage Decomposition of Word-Problem Solving and Mechanistic Fragility in LLM Math Reasoning
This paper analyzes how large language models solve math word problems and identifies fragile mechanistic stages, offering insight into their reasoning reliability.
This paper analyzes how large language models solve math word problems and identifies fragile mechanistic stages, offering insight into their reasoning reliability.
Auto-scored: A research paper on LLM math reasoning fragility poses only minor framing-level risk.