推理模型越擅长深度思考，基础能力反而越差，自适应推理可缓解这一问题。
Trade-offs in Large Reasoning Models: An Empirical Analysis of Deliberative and Adaptive Reasoning over Foundational Capabilities
- 通过零思考、少思考等模式动态调整推理强度
- 深度思考使模型帮助性下降30%，有害性上升25%
- 适合需要平衡性能与成本的工程应用
近期大型推理模型（LRMs）如OpenAI的o1/o3和DeepSeek-R1在特定推理任务中展现出类人深度思考与长链推理能力。然而，我们在多个模型家族（DeepSeek、Qwen、LLaMA）和规模（7B至32B）上进行系统评估发现，获取此类深度推理能力会显著削弱基础能力，包括帮助性下降30%、有害性上升25%，并导致推理成本大幅增加。重要的是，我们证明采用零思考、少思考、摘要思考等自适应推理模式能有效缓解这些缺陷。实证结果强调了开发具备动态分配计算资源能力的多功能推理模型的紧迫性。
原文摘要 · Abstract (English)
Recent advancements in Large Reasoning Models (LRMs), such as OpenAI's o1/o3 and DeepSeek-R1, have demonstrated remarkable performance in specialized reasoning tasks through human-like deliberative thinking and long chain-of-thought reasoning. However, our systematic evaluation across various model families (DeepSeek, Qwen, and LLaMA) and scales (7B to 32B) reveals that acquiring these deliberative reasoning capabilities significantly reduces the foundational capabilities of LRMs, including notable declines in helpfulness and harmlessness, alongside substantially increased inference costs. Importantly, we demonstrate that adaptive reasoning -- employing modes like Zero-Thinking, Less-Thinking, and Summary-Thinking -- can effectively alleviate these drawbacks. Our empirical insights underline the critical need for developing more versatile LRMs capable of dynamically allocating inference-time compute according to specific task characteristics.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。