Teaching LLM to Reason: Reinforcement Learning from Algorithmic Problems without Code
This addresses the issue of overfitting in LLM reasoning training for researchers and practitioners, offering an incremental improvement over existing methods.
The paper tackles the problem of improving LLM reasoning by reducing overfitting to algorithmic patterns in code-based training, proposing TeaR, which uses curated data and reinforcement learning to guide reasoning paths, resulting in performance improvements such as 35.9% on Qwen2.5-7B and 5.9% on R1-Distilled-7B across 17 benchmarks.
Enhancing reasoning capabilities remains a central focus in the LLM reasearch community. A promising direction involves requiring models to simulate code execution step-by-step to derive outputs for given inputs. However, as code is often designed for large-scale systems, direct application leads to over-reliance on complex data structures and algorithms, even for simple cases, resulting in overfitting to algorithmic patterns rather than core reasoning structures. To address this, we propose TeaR, which aims at teaching LLMs to reason better. TeaR leverages careful data curation and reinforcement learning to guide models in discovering optimal reasoning paths through code-related tasks, thereby improving general reasoning abilities. We conduct extensive experiments using two base models and three long-CoT distillation models, with model sizes ranging from 1.5 billion to 32 billion parameters, and across 17 benchmarks spanning Math, Knowledge, Code, and Logical Reasoning. The results consistently show significant performance improvements. Notably, TeaR achieves a 35.9% improvement on Qwen2.5-7B and 5.9% on R1-Distilled-7B.