通过减法思考：基于置信度的对比解码用于大语言模型推理
Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning
AI总结：
通过置信度驱动的对比解码方法提升大语言模型推理可靠性，减少输出长度并提高准确性。
摘要：
近期关于大语言模型（LLM）推理的测试时间扩展工作通常假设在推理时间内分配更多计算资源能均匀提高正确性。然而，先前研究显示推理不确定性高度局部化：一小部分低置信度的token disproportionately 贡献于推理错误和不必要的输出扩展。受此观察启发，我们提出了通过减法思考（Thinking by Subtraction），一种基于置信度的对比解码方法，通过有针对性的token级干预提高推理可靠性。我们的方法，置信度驱动对比解码（Confidence-Driven Contrastive Decoding），在解码过程中检测低置信度的token，并在这些位置进行选择性干预。它通过将高置信度token替换为最小占位符构建对比参考，并通过在低置信度位置减去此参考分布来优化预测。实验表明，CCD在数学推理基准上显著提高了准确性，同时大幅减少了输出长度，且具有最小的KV缓存开销。作为一种无需训练的方法，CCD通过有针对性的低置信度干预增强推理可靠性，而无需计算冗余。我们的代码将在：https://github.com/bolo-web/CCD 上提供。
英文摘要：
Recent work on test-time scaling for large language model (LLM) reasoning typically assumes that allocating more inference-time computation uniformly improves correctness. However, prior studies show that reasoning uncertainty is highly localized: a small subset of low-confidence tokens disproportionately contributes to reasoning errors and unnecessary output expansion. Motivated by this observation, we propose Thinking by Subtraction, a confidence-driven contrastive decoding approach that improves reasoning reliability through targeted token-level intervention. Our method, Confidence-Driven Contrastive Decoding, detects low-confidence tokens during decoding and intervenes selectively at these positions. It constructs a contrastive reference by replacing high-confidence tokens with minimal placeholders, and refines predictions by subtracting this reference distribution at low-confidence locations. Experiments show that CCD significantly improves accuracy across mathematical reasoning benchmarks while substantially reducing output length, with minimal KV-cache overhead. As a training-free method, CCD enhances reasoning reliability through targeted low-confidence intervention without computational redundancy. Our code will be made available at: https://github.com/bolo-web/CCD.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
