ReasonBridge: Efficient Reasoning Transfer from Closed to Open-Source Language Models

TL;DR

ReasonBridge presents a hierarchical knowledge distillation method that significantly enhances reasoning abilities in open-source language models, narrowing the performance gap with closed-source models through a novel dataset and efficient transfer techniques.

Contribution

This work introduces a hierarchical distillation framework and a curated reasoning dataset to transfer reasoning skills from closed to open-source models effectively.

Findings

Up to 23% improvement in reasoning tasks for open-source models

Qwen2.5-14B surpasses Claude-Sonnet3.5 on MATH500

Method generalizes across domains and architectures

Abstract

Recent advancements in Large Language Models (LLMs) have revealed a significant performance gap between closed-source and open-source models, particularly in tasks requiring complex reasoning and precise instruction following. This paper introduces ReasonBridge, a methodology that efficiently transfers reasoning capabilities from powerful closed-source to open-source models through a novel hierarchical knowledge distillation framework. We develop a tailored dataset Reason1K with only 1,000 carefully curated reasoning traces emphasizing difficulty, diversity, and quality. These traces are filtered from across multiple domains using a structured multi-criteria selection algorithm. Our transfer learning approach incorporates: (1) a hierarchical distillation process capturing both strategic abstraction and tactical implementation patterns, (2) a sparse reasoning-focused adapter architecture requiring only 0.3% additional trainable parameters, and (3) a test-time compute scaling mechanism using guided inference interventions. Comprehensive evaluations demonstrate that ReasonBridge improves reasoning capabilities in open-source models by up to 23% on benchmark tasks, significantly narrowing the gap with closed-source models. Notably, the enhanced Qwen2.5-14B outperforms Claude-Sonnet3.5 on MATH500 and matches its performance on competition-level AIME problems. Our methodology generalizes effectively across diverse reasoning domains and model architectures, establishing a sample-efficient approach to reasoning enhancement for instruction following.