Debiasing LLMs by Masking Unfairness-Driving Attention Heads (2026.findings-acl)
Copied to clipboard
Tingxu Han, Wei Song, Ziqi Ding, Ziming Li, Chunrong Fang, Yuekang Li, Dongfang Liu, Zhenyu Chen, Zhenting Wang
| Challenge: | Existing work probes when biased outputs appear, but gives little insight into the mechanisms that generate them, leaving existing mitigations largely fragile. |
| Approach: | They propose a lightweight debiasing framework that detects bias heads and selectively masks only those heads that activate under DA and CoT. |
| Outcome: | The proposed framework reduces unfairness by 391.9%- 534.5% in both one- and two-turn dialogues. |
Similar Papers
FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation Steering (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing prompt-based debiasing methods exhibit instability due to sensitivity to prompt changes . fine-tuning-based techniques incur substantial computational overhead and catastrophic forgetting . |
| Approach: | They propose a debiasing framework that encodes fairness-related features into separable directions in the hidden activation space. |
| Outcome: | The proposed framework performs inference-time debiasing without requiring retraining or prompt design . it detects bias signatures in activations and then computes debiased steering vectors . the proposed framework is available to download in the u.s. |
DeFrame: Debiasing Large Language Models Against Framing Effects (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing debiasing methods improve overall fairness, but fail to reduce framing-induced disparities. |
| Approach: | They propose a framing-aware debiasing method that encourages LLMs to be more consistent across frams. |
| Outcome: | The proposed method reduces overall bias and improves robustness against framing disparities, enabling LLMs to produce fairer and more consistent responses. |
Rethinking Prompt-based Debiasing in Large Language Model (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing prompt-based methods for debiasing are often superficial and lack a thorough understanding of complex bias concepts. |
| Approach: | They analyze a BBQ and stereoSet benchmarks to examine the assumption that large language models understand biases. |
| Outcome: | The proposed model misclassified 90% of unbiased content as biased despite high accuracy on BBQ dataset . the proposed model may have been flawed in previous attempts to debiase . |
“Thinking” Fair and Slow: On the Efficacy of Structured Prompts for Debiasing Language Models (2024.emnlp-main)
Copied to clipboard
Shaz Furniturewala, Surgan Jandial, Abhinav Java, Pragyan Banerjee, Simra Shahid, Sumit Bhatia, Kokil Jaidka
| Challenge: | Existing debiasing techniques are typically training-based or require access to the model’s internals and output distributions, so they are inaccessible to end-users looking to adapt LLM outputs for their particular needs. |
| Approach: | They propose a system-based iterative framework that uses System 2 thinking processes to induce logical, reflective, and critical text generation with single, multi-step, instruction, and role-based variants. |
| Outcome: | The proposed framework significantly improves over other frameworks demonstrating lower mean bias in the outputs with competitive performance on the downstream tasks. |
Cognitive Bias in Decision-Making with LLMs (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models inherit societal biases against protected groups and can be subject to functionally resembling cognitive bias. |
| Approach: | They propose a framework to uncover, evaluate, and mitigate cognitive bias in large language models by using a dataset containing 13,465 prompts to evaluate LLM decisions on different cognitive biases. |
| Outcome: | The proposed framework uncovers, evaluates, and mitigates cognitive bias in large language models, particularly in high-stakes decision-making tasks. |
DeCoT: Debiasing Chain-of-Thought for Knowledge-Intensive Tasks in Large Language Models via Causal Intervention (2024.acl-long)
Copied to clipboard
| Challenge: | In large language models, external knowledge is required to augment their internal knowledge through prompts, but this does not guarantee that LLMs can identify and use relevant information in the prompts to conduct chain-of-thought reasoning. |
| Approach: | They propose a structural causal model to formally explain the internal knowledge bias of large language models (LLMs) they review the chain-of-thought (CoT) prompting from a causal perspective and find that biased information from pretrained models can impair LLMs’ reasoning abilities. |
| Outcome: | The proposed model enables more accurate CoT reasoning and enhances LLM generation on knowledge-intensive tasks. |
Toward Inclusive Language Models: Sparsity-Driven Calibration for Systematic and Interpretable Mitigation of Social Biases in LLMs (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a new method to mitigate stereotypical bias in large language models is needed . inherent biases from training on vast Internet datasets can amplify harmful stereotypes . |
| Approach: | They propose a method to identify stereotypical bias in decoder-only transformer models . they apply a localization mechanism that correlates internal activations with a new Context Influence score . |
| Outcome: | The proposed method reduces stereotypical biases on BBQ, StereoSet, and CrowS-Pairs while improving reasoning performance on MMLU by 10%. |
Can We Instruct LLMs to Compensate for Position Bias? (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies reveal that position bias in large language models (LLMs) leads to difficulty in accessing information retrieved from the retriever. |
| Approach: | They propose to direct LLMs to allocate more attention towards a selected segment of the context through prompting. |
| Outcome: | The proposed approach improves the performance of large language models by promoting instruction with an exact document index. |
Fine-tuning LLMs with Cross-Attention-based Weight Decay for Bias Mitigation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) excel in natural language processing tasks but often propagate societal biases from their training data, leading to discriminatory outputs. |
| Approach: | They propose a method that modifies the LLM architecture to mitigate bias by adjusting the attention weights of sensitive tokens. |
| Outcome: | The proposed method can handle multiple sensitive attributes and does not require full knowledge of sensitive tokens presented in the dataset. |
Beyond Performance: Quantifying and Mitigating Label Bias in LLMs (2024.naacl-long)
Copied to clipboard
| Challenge: | Large language models exhibit undesirable preference toward predicting certain answers over others, despite their adaptability to diverse tasks. |
| Approach: | They propose a label bias calibration method that outperforms recent calibration approaches for improving performance and mitigating label bias. |
| Outcome: | The proposed method outperforms calibration approaches for improving performance and mitigating label bias. |