COFT: Counterfactual-Conformal Decoding for Fair Chain-of-Thought Reasoning in Large Language Models
Quick Take
COFT (Chain of Fair Thought) is a novel decoding method that mitigates biases in large language models by applying token-level fairness control during decoding. It achieves a 30-55% reduction in bias metrics across six models without retraining, while maintaining task utility and language quality, with only a modest computational overhead.
Key Points
- COFT applies token-level fairness control at decode time for large language models.
- Reduces bias metrics by 30-55% while preserving task utility and language quality.
- Evaluated across six models with significant bias reduction and negligible utility loss.
- Computational overhead is modest, equivalent to one additional cached forward pass.
- No retraining or auxiliary classifiers are required for implementation.
Article Content
From source RSS / original summaryarXiv:2605. 30641v1 Announce Type: new Abstract: 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.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
The REFLECT benchmark reveals that current LLM judges are unreliable, achieving below 55% accuracy in evaluating reasoning and evidence use, highlighting the need for improved evaluation methods for deep research agents.