Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning
This work addresses a critical problem for AI researchers and practitioners by revealing limitations in current tuning methods for LLMs, suggesting it is incremental in highlighting specific transfer issues rather than proposing a new paradigm.
The study investigated whether improvements in math reasoning for large language models (LLMs) transfer to broader problem-solving abilities, finding that most models fail to generalize gains to other domains like scientific QA and coding, with RL-tuned models showing better transfer than SFT-tuned ones.
Math reasoning has become the poster child of progress in large language models (LLMs), with new models rapidly surpassing human-level performance on benchmarks like MATH and AIME. But as math leaderboards improve week by week, it is worth asking: do these gains reflect broader problem-solving ability or just narrow overfitting? To answer this question, we evaluate over 20 open-weight reasoning-tuned models across a broad suite of tasks, including math, scientific QA, agent planning, coding, and standard instruction-following. We surprisingly find that most models that succeed in math fail to transfer their gains to other domains. To rigorously study this phenomenon, we conduct controlled experiments on Qwen3-14B models using math-only data but different tuning methods. We find that reinforcement learning (RL)-tuned models generalize well across domains, while supervised fine-tuning (SFT)-tuned models often forget general capabilities. Latent-space representation and token-space distribution shift analyses reveal that SFT induces substantial representation and output drift, while RL preserves general-domain structure. Our results suggest a need to rethink standard post-training recipes, particularly the reliance on SFT-distilled data for advancing reasoning models.