Reasoning Pattern Alignment Merging for Adaptive Reasoning

TL;DR

This paper introduces RPAM, a layer-wise model merging framework that adaptively combines long and short chain-of-thought reasoning models, reducing inference costs while maintaining performance on reasoning tasks.

Contribution

The paper proposes a novel, training-free model merging method based on feature alignment for adaptive reasoning, improving efficiency in large reasoning models.

Findings

RPAM significantly reduces inference cost.

RPAM maintains strong reasoning performance.

RPAM adapts reasoning patterns effectively.

Abstract

Recent large reasoning models (LRMs) have made substantial progress in complex reasoning tasks, yet they often generate lengthy reasoning paths for every query, incurring unnecessary computation and latency. Existing speed-up approaches typically rely on retraining the model or designing sophisticated prompting, which are either prohibitively expensive or highly sensitive to the input and prompt formulation. In this work, we study model merging as a lightweight alternative for efficient reasoning: by combining a long chain-of-thought (Long-CoT) reasoning model with a Short-CoT instruction model, we obtain an adaptive reasoner without training from scratch or requiring large-scale additional data. Building on this idea, we propose Reasoning Pattern Alignment Merging (RPAM), a layer-wise model merging framework based on feature alignment to facilitate query-adaptive reasoning. RPAM first constructs a small pattern-labeled calibration set that assigns each query an appropriate reasoning pattern. It then optimizes layer-wise merging coefficients by aligning the merged model's intermediate representations with those of the selected model, while a contrastive objective explicitly pushes them away from the non-selected model. Experiments on seven widely used reasoning benchmarks show that RPAM substantially reduces inference cost while maintaining strong performance. Upon article acceptance, we will provide open-source code to reproduce experiments for RPAM.