Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning
Authors:
- Lexiang TangLTLexiang Tang
- Weihao GaoWGWeihao Gao
- Bingchen ZhaoBZBingchen Zhao
- Lu MaLMLu Ma
- Qiao JinQJQiao Jin
- Bang YangBYBang Yang
- Yuexian ZouYZYuexian Zou
Preprint
Abstract
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.Keywords
All Related Versions
This publication has 6 references indexed in Scilit:
- Not All Tokens and Heads Are Equally Important: Dual-Level Attention Intervention for Hallucination MitigationProceedings of the AAAI Conference on Artificial Intelligence, 2026
- Entropy-Gated Branching for Efficient Test-Time Reasoning2026
- Confidence Improves Self-Consistency in LLMs2025
- Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decoding2024
- Enhancing Contextual Understanding in Large Language Models through Contrastive Decoding2024
- Self-Refine: Iterative Refinement with Self-Feedback2023