COFT: Counterfactual-Conformal Decoding for Fair Chain-of-Thought Reasoning in Large Language Models
On the map: LLM reasoning → Prompting and chain-of-thought reasoning
Abstract (arXiv)
Large language models (LLMs) can reveal and amplify societal biases during chain-of-thought (CoT) generation. We present COFT (Chain of Fair Thought), a training-free decoding method that applies token-level fairness control at decode time, with distribution-free marginal validity guarantees (under exchangeability) for any frozen causal language model. COFT operates in three stages. First, it creates a masked counterfactual prompt by replacing sensitive spans with neutral tokens. Second, it compares the factual and masked logit distributions through lightweight logit fusion to attenuate attribute-driven biases. Third, it uses dual-branch split-conformal calibration to certify per-step candidate token sets at a user-chosen risk level. We evaluate COFT across six models and multiple bias benchmarks. Our method reduces standard bias metrics by 30-55% (median 38%) while preserving task utility and language quality. Reasoning accuracies remain unchanged within run-to-run noise margins. The computational overhead is modest, equivalent to one additional cached forward pass (<=11%). COFT offers a clear, auditable path to safer CoT generation with significant bias reduction, negligible utility loss, and no requirement for retraining, auxiliary classifiers, or weight access.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 17 of its 60 references that are in the map. First posted on arXiv 28 May 2026.
Builds on
17 of 60 references.
References · 17
- 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
- 2021Training Verifiers to Solve Math Word Problemscited by 9,743
- 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
- 2023Mistral 7Bcited by 3,065
- 2016Pointer Sentinel Mixture Modelscited by 2,992
- 2024Qwen2 Technical Reportcited by 2,076
- 2024Mixtral of Expertscited by 1,785
- 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debatecited by 995
- 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 951
- 2021A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantificationcited by 777
- 2023SLiC-HF: Sequence Likelihood Calibration with Human Feedbackcited by 395
- 2024Large language model validity via enhanced conformal prediction methodscited by 91
5 more
- 2024MC-CoT: A Modular Collaborative CoT Framework for Zero-shot Medical-VQA with LLM and MLLM Integrationcited by 21
- 2026Power-SMC: Low-Latency Sequence-Level Power Sampling for Training-Free LLM Reasoningcited by 17
- 2025Conformal Arbitrage: Risk-Controlled Balancing of Competing Objectives in Language Modelscited by 10
- 2025FACTER: Fairness-Aware Conformal Thresholding and Prompt Engineering for Enabling Fair LLM-Based Recommender Systemscited by 6
- 2026Proof-of-Perception: Certified Tool-Using Multimodal Reasoning with Compositional Conformal Guaranteescited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Reasoning Towards Fairness: Mitigating Bias in Language Models through Reasoning-Guided Fine-Tuning
- 2025CounterBench: Evaluating and Improving Counterfactual Reasoning in Large Language Models
- 2025FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning
- 2025BiasFilter: An Inference-Time Debiasing Framework for Large Language Models
- 2026SyncThink: A Training-Free Strategy to Align Inference Termination with Reasoning Saturation
- 2025Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs
- 2025Rationale-Enhanced Decoding for Multi-modal Chain-of-Thought
- 2024FairCoT: Enhancing Fairness in Text-to-Image Generation via Chain of Thought Reasoning with Multimodal Large Language Models