通过弹性推理实现可扩展的思维链

Scalable Chain of Thoughts via Elastic Reasoning

May 8, 2025

作者: Yuhui Xu, Hanze Dong, Lei Wang, Doyen Sahoo, Junnan Li, Caiming Xiong

cs.AI

摘要

大型推理模型（LRMs）通过生成扩展的思维链（CoT）在复杂任务上取得了显著进展。然而，其不受控制的输出长度在实际部署中带来了重大挑战，特别是在推理时对令牌数量、延迟或计算资源的严格限制下。我们提出了弹性推理（Elastic Reasoning），这是一种新颖的可扩展思维链框架，明确将推理分为两个阶段——思考阶段和解答阶段，并各自独立分配预算。在测试时，弹性推理优先保证解答片段的完整性，在资源紧张的情况下显著提高了可靠性。为了训练出能够适应思考过程被截断的模型，我们引入了一种轻量级的预算约束展开策略，并将其整合到GRPO中，该策略教导模型在思考过程被中断时自适应地进行推理，并能有效泛化到未见过的预算约束，无需额外训练。在数学（AIME、MATH500）和编程（LiveCodeBench、Codeforces）基准测试中的实证结果表明，弹性推理在严格的预算约束下表现稳健，同时训练成本显著低于基线方法。值得注意的是，即使在无约束环境下，我们的方法也能生成更简洁高效的推理。弹性推理为大规模可控推理这一紧迫挑战提供了一个原则性和实用性的解决方案。

English

Large reasoning models (LRMs) have achieved remarkable progress on complex tasks by generating extended chains of thought (CoT). However, their uncontrolled output lengths pose significant challenges for real-world deployment, where inference-time budgets on tokens, latency, or compute are strictly constrained. We propose Elastic Reasoning, a novel framework for scalable chain of thoughts that explicitly separates reasoning into two phases--thinking and solution--with independently allocated budgets. At test time, Elastic Reasoning prioritize that completeness of solution segments, significantly improving reliability under tight resource constraints. To train models that are robust to truncated thinking, we introduce a lightweight budget-constrained rollout strategy, integrated into GRPO, which teaches the model to reason adaptively when the thinking process is cut short and generalizes effectively to unseen budget constraints without additional training. Empirical results on mathematical (AIME, MATH500) and programming (LiveCodeBench, Codeforces) benchmarks demonstrate that Elastic Reasoning performs robustly under strict budget constraints, while incurring significantly lower training cost than baseline methods. Remarkably, our approach also produces more concise and efficient reasoning even in unconstrained settings. Elastic Reasoning offers a principled and practical solution to the pressing challenge of controllable reasoning at scale.