让专业模型具备深度推理能力，且不损失原有专长。
ReasonAny: Incorporating Reasoning Capability to Any Model via Simple and Effective Model Merging
- 通过对比梯度识别，精准定位推理能力所在参数区域。
- 在安全、生物医学、金融领域均显著超越现有方法。
- 适合希望低成本提升模型推理能力的研究者和开发者。
大型推理模型（LRMs）凭借长链式思维推理取得了显著成果。然而，为领域专用模型赋予此类推理能力（即“推理+X”）仍是重大挑战。尽管模型合并提供了一种无需训练的解决方案，但现有方法常导致性能崩溃：既削弱推理深度，又损害领域专用能力。我们发现一个反直觉现象：推理能力主要集中在梯度敏感度低的参数区域，而非通常认为的高幅值参数。基于此，我们提出ReasonAny，一种新型合并框架，通过对比梯度识别有效解决推理与领域性能的冲突。在安全、生物医学、金融等领域的实验表明，ReasonAny能有效融合“推理+X”能力，显著优于当前最优基线，同时保持强大的推理表现。
原文摘要 · Abstract (English)
Large Reasoning Models (LRMs) with long chain-of-thought reasoning have recently achieved remarkable success. Yet, equipping domain-specialized models with such reasoning capabilities, referred to as "Reasoning + X", remains a significant challenge. While model merging offers a promising training-free solution, existing methods often suffer from a destructive performance collapse: existing methods tend to both weaken reasoning depth and compromise domain-specific utility. Interestingly, we identify a counter-intuitive phenomenon underlying this failure: reasoning ability predominantly resides in parameter regions with low gradient sensitivity, contrary to the common assumption that domain capabilities correspond to high-magnitude parameters. Motivated by this insight, we propose ReasonAny, a novel merging framework that resolves the reasoning-domain performance collapse through Contrastive Gradient Identification. Experiments across safety, biomedicine, and finance domains show that ReasonAny effectively synthesizes "Reasoning + X" capabilities, significantly outperforming state-of-the-art baselines while retaining robust reasoning performance.
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。