ReasonAny: Incorporating Reasoning Capability to Any Model via Simple and Effective Model Merging

TL;DR

ReasonAny is a novel model merging framework that effectively combines reasoning capabilities with domain-specific models without training, overcoming performance collapse issues by identifying parameter regions crucial for reasoning.

Contribution

It introduces Contrastive Gradient Identification to improve model merging, enabling domain models to gain reasoning skills without sacrificing domain performance.

Findings

Outperforms state-of-the-art baselines in multiple domains

Retains strong reasoning capabilities after merging

Successfully combines reasoning with domain expertise without training

Abstract

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.