Short Chains, Deep Thoughts: Balancing Reasoning Efficiency and Intra-Segment Capability via Split-Merge Optimization

2026-02-03

AI Summary

This work addresses the high latency and computational overhead of large reasoning models caused by structural redundancy in generating long reasoning chains. To mitigate this, the authors propose CoSMo, a novel framework that integrates consistency-guided dynamic merge-and-split optimization, structure-aligned reinforcement learning, and segment-level computational budget constraints to streamline reasoning chains while preserving logical completeness. Experimental results demonstrate that CoSMo consistently improves accuracy by an average of 3.3 percentage points across multiple benchmarks and backbone models, while reducing the number of reasoning segments used by 28.7%, thereby achieving a significant balance between efficiency and expressive reasoning capability.

Technology Category

• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

While Large Reasoning Models (LRMs) have demonstrated impressive capabilities in solving complex tasks through the generation of long reasoning chains, this reliance on verbose generation results in significant latency and computational overhead. To address these challenges, we propose \textbf{CoSMo} (\textbf{Co}nsistency-Guided \textbf{S}plit-\textbf{M}erge \textbf{O}ptimization), a framework designed to eliminate structural redundancy rather than indiscriminately restricting token volume. Specifically, CoSMo utilizes a split-merge algorithm that dynamically refines reasoning chains by merging redundant segments and splitting logical gaps to ensure coherence. We then employ structure-aligned reinforcement learning with a novel segment-level budget to supervise the model in maintaining efficient reasoning structures throughout training. Extensive experiments across multiple benchmarks and backbones demonstrate that CoSMo achieves superior performance, improving accuracy by \textbf{3.3} points while reducing segment usage by \textbf{28.7\%} on average compared to reasoning efficiency baselines.

Problem

• Large Reasoning Models
• reasoning efficiency
• computational overhead
• reasoning chains
• latency

Innovation

• Split-Merge Optimization
• Reasoning Efficiency
• Structural Redundancy
• Segment-Level Budget
• Consistency-Guided Reinforcement Learning