Reasoning Curriculum: Bootstrapping Broad LLM Reasoning from Math
This work addresses the challenge of general reasoning in LLMs for AI applications, offering a compact and easy-to-adopt method, though it is incremental as it builds on existing reinforcement learning approaches.
The paper tackles the problem of eliciting broad reasoning skills in large language models by proposing a two-stage curriculum that first develops reasoning in math and then transfers it to other domains, resulting in consistent performance gains across a multi-domain evaluation suite.
Reinforcement learning (RL) can elicit strong reasoning in large language models (LLMs), yet most open efforts focus on math and code. We propose Reasoning Curriculum, a simple two-stage curriculum that first elicits reasoning skills in pretraining-aligned domains such as math, then adapts and refines these skills across other domains via joint RL. Stage 1 performs a brief cold start and then math-only RL with verifiable rewards to develop reasoning skills. Stage 2 runs joint RL on mixed-domain data to transfer and consolidate these skills. The curriculum is minimal and backbone-agnostic, requiring no specialized reward models beyond standard verifiability checks. Evaluated on Qwen3-4B and Llama-3.1-8B over a multi-domain suite, reasoning curriculum yields consistent gains. Ablations and a cognitive-skill analysis indicate that both stages are necessary and that math-first elicitation increases cognitive behaviors important for solving complex problems. Reasoning Curriculum provides a compact, easy-to-adopt recipe for general reasoning.