FINEREASON: Evaluating and Improving LLMs' Deliberate Reasoning through Reflective Puzzle Solving
This addresses the need for better evaluation of reasoning processes in AI, though it is incremental as it builds on existing benchmarks and methods.
The paper tackles the problem of evaluating large language models' deliberate reasoning by introducing FINEREASON, a logic-puzzle benchmark that assesses intermediate steps, and shows that training on this data improves math reasoning performance by up to 5.1% on GSM8K.
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 fine-grained 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. We show that models trained on our state checking and transition data demonstrate gains in math reasoning by up to 5.1% on GSM8K.