Go to ACL ARR 2025 February homepageFineReason: Evaluating and Improving LLMs' Deliberate Reasoning through Reflective Puzzle Solving
Abstract: Many challenging reasoning tasks require not just rapid, intuitive responses, but a more deliberate, multi-step approach. Recent progress in large language models (LLMs) highlights an important shift from the "System 1" way of quick reactions to the "System 2" style of reflection-and-correction problem solving. However, current benchmarks heavily rely on the final-answer accuracy, leaving much of a model's intermediate reasoning steps unexamined. This fails to assess the model's ability to reflect and rectify mistakes within the reasoning process. To bridge this gap, we introduce FineReason, a logic-puzzle benchmark for granular evaluation of LLMs' reasoning capabilities. Each puzzle can be decomposed into atomic steps, making it ideal for rigorous validation of intermediate correctness. Building on this, we introduce two tasks: state checking, and state transition, for a comprehensive evaluation of how models assess the current situation and plan the next move. To support broader research, we also provide a puzzle training set aimed at enhancing performance on general mathematical tasks. To support broader research, we also provide a puzzle training set aimed at enhancing performance on general mathematical tasks. We show that models trained on our state checking and transition data demonstrate gains in overall math reasoning from 82.3% to 87.4% on GSM8K using the DeepSeek-R1-Distill-Qwen-7B model.
Paper Type: Long
Research Area: Resources and Evaluation
Research Area Keywords: Reasoning, Large Langauge Model, Evaluation
Contribution Types: Model analysis & interpretability, NLP engineering experiment, Data resources, Data analysis
Languages Studied: English
Submission Number: 6824
Loading