LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation

2026-02-15

AI Summary

This work addresses the challenges of “under-thinking” (insufficient reasoning depth) and “over-thinking” (redundant, inefficient exploration) in code generation by proposing a lightweight, logit-level control mechanism that iteratively refines reasoning steps to achieve efficient and structured code synthesis. The approach integrates Logits Preference Decoding to guide token selection, a logits-ranking-based path selection strategy to prioritize high-quality reasoning trajectories, and a thought-aggregation mechanism to balance reasoning depth with computational efficiency. Experimental results demonstrate that the proposed method significantly enhances both the quality of reasoning chains and generation efficiency, outperforming existing baselines across multiple code generation benchmarks.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• 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

Code generation remains a challenging task that requires precise and structured reasoning. Existing Test Time Scaling (TTS) methods, including structured tree search, have made progress in exploring reasoning paths but still face two major challenges: (1) underthinking, where reasoning chains tend to be shallow and fail to capture the full complexity of problems; and (2) overthinking, where overly verbose reasoning leads to inefficiency and increased computational costs. To address these issues, we propose LogitsCoder, a novel framework that enhances chain-of-thought reasoning through lightweight, logit-level control mechanisms for code generation. LogitsCoder iteratively generates and refines reasoning steps by first steering token selection toward statistically preferred patterns via Logits Preference Decoding, then selecting and aggregating diverse reasoning paths using Logits Rank Based Path Selection and Thoughts Aggregation. This results in coherent and effective reasoning chains that balance depth and efficiency. Extensive experiments demonstrate that LogitsCoder produces more efficient and higher-quality reasoning chains, leading to superior code generation performance compared to baseline methods.

Problem

• code generation
• chain-of-thought
• reasoning paths
• underthinking
• overthinking

Innovation

• Logits Preference Decoding
• Chain-of-Thought
• Code Generation
• Test Time Scaling
• Path Selection