DeepCompress: A Dual Reward Strategy for Dynamically Exploring and Compressing Reasoning Chains
This work addresses efficiency and accuracy trade-offs in reasoning models for AI applications, representing an incremental improvement over existing methods.
The paper tackled the problem of cognitive inefficiencies in Large Reasoning Models, such as overthinking simple problems and underthinking complex ones, by introducing DeepCompress, a framework that dynamically adjusts reasoning chain lengths based on problem difficulty, resulting in superior accuracy and significant token efficiency improvements on mathematical benchmarks.
Large Reasoning Models (LRMs) have demonstrated impressive capabilities but suffer from cognitive inefficiencies like ``overthinking'' simple problems and ``underthinking'' complex ones. While existing methods that use supervised fine-tuning~(SFT) or reinforcement learning~(RL) with token-length rewards can improve efficiency, they often do so at the cost of accuracy. This paper introduces \textbf{DeepCompress}, a novel framework that simultaneously enhances both the accuracy and efficiency of LRMs. We challenge the prevailing approach of consistently favoring shorter reasoning paths, showing that longer responses can contain a broader range of correct solutions for difficult problems. DeepCompress employs an adaptive length reward mechanism that dynamically classifies problems as ``Simple'' or ``Hard'' in real-time based on the model's evolving capability. It encourages shorter, more efficient reasoning for ``Simple'' problems while promoting longer, more exploratory thought chains for ``Hard'' problems. This dual-reward strategy enables the model to autonomously adjust its Chain-of-Thought (CoT) length, compressing reasoning for well-mastered problems and extending it for those it finds challenging. Experimental results on challenging mathematical benchmarks show that DeepCompress consistently outperforms baseline methods, achieving superior accuracy while significantly improving token efficiency.