BiasGym: Fantastic LLM Biases and How to Find (and Remove) Them

• URL: http://arxiv.org/abs/2508.08855v2
• Date: Thu, 14 Aug 2025 17:57:53 GMT
• Title: BiasGym: Fantastic LLM Biases and How to Find (and Remove) Them
• Authors: Sekh Mainul Islam, Nadav Borenstein, Siddhesh Milind Pawar, Haeun Yu, Arnav Arora, Isabelle Augenstein,
• Abstract summary: BiasGym is a framework for reliably injecting, analyzing, and mitigating conceptual associations within Large Language Models (LLMs)<n>BiasGym consists of two components: BiasInject, which injects specific biases into the model via token-based fine-tuning while keeping the model frozen, and BiasScope, which leverages these injected signals to identify and steer the components responsible for biased behavior.<n>Our method enables consistent bias elicitation for mechanistic analysis, supports targeted debiasing without degrading performance on downstream tasks, and generalizes to biases unseen during token-based fine-tuning.
• Score: 38.80876158025777
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Understanding biases and stereotypes encoded in the weights of Large Language Models (LLMs) is crucial for developing effective mitigation strategies. Biased behaviour is often subtle and non-trivial to isolate, even when deliberately elicited, making systematic analysis and debiasing particularly challenging. To address this, we introduce BiasGym, a simple, cost-effective, and generalizable framework for reliably injecting, analyzing, and mitigating conceptual associations within LLMs. BiasGym consists of two components: BiasInject, which injects specific biases into the model via token-based fine-tuning while keeping the model frozen, and BiasScope, which leverages these injected signals to identify and steer the components responsible for biased behavior. Our method enables consistent bias elicitation for mechanistic analysis, supports targeted debiasing without degrading performance on downstream tasks, and generalizes to biases unseen during token-based fine-tuning. We demonstrate the effectiveness of BiasGym in reducing real-world stereotypes (e.g., people from Italy being `reckless drivers') and in probing fictional associations (e.g., people from a fictional country having `blue skin'), showing its utility for both safety interventions and interpretability research.

Related papers

• Making Bias Non-Predictive: Training Robust LLM Judges via Reinforcement Learning [91.8584139564909]
  Large language models (LLMs) increasingly serve as automated judges, yet they remain susceptible to cognitive biases.<n>We propose Epistemic Independence Training (EIT), a reinforcement learning framework grounded in a key principle.<n>EIT operationalizes this through a balanced conflict strategy where bias signals are equally likely to support correct and incorrect answers.
  arXiv  Detail & Related papers  (2026-02-02T01:43:48Z)
• No Free Lunch in Language Model Bias Mitigation? Targeted Bias Reduction Can Exacerbate Unmitigated LLM Biases [6.184434080778806]
  Large Language Models (LLMs) inherit societal biases from their training data, potentially leading to harmful or unfair outputs.<n>We study four bias mitigation techniques applied across ten models from seven model families, and we explore racial, religious, profession- and gender-related biases.<n>Our results consistently show that while targeted mitigation can sometimes reduce bias in the intended dimension, it frequently leads to unintended and often negative consequences in others.
  arXiv  Detail & Related papers  (2025-11-23T22:21:18Z)
• Explicit vs. Implicit: Investigating Social Bias in Large Language Models through Self-Reflection [18.625071242029936]
  Large Language Models (LLMs) have been shown to exhibit various biases and stereotypes in their generated content.<n>This paper presents a systematic framework to investigate and compare explicit and implicit biases in LLMs.
  arXiv  Detail & Related papers  (2025-01-04T14:08:52Z)
• Towards Resource Efficient and Interpretable Bias Mitigation in Large Language Models [1.787433808079955]
  Large language models (LLMs) have been observed to perpetuate unwanted biases in training data.<n>In this paper, we mitigate bias by leveraging small biased and anti-biased expert models to obtain a debiasing signal.<n> Experiments on mitigating gender, race, and religion biases show a reduction in bias on several local and global bias metrics.
  arXiv  Detail & Related papers  (2024-12-02T16:56:08Z)
• Identifying and Mitigating Social Bias Knowledge in Language Models [52.52955281662332]
  We propose a novel debiasing approach, Fairness Stamp (FAST), which enables fine-grained calibration of individual social biases.<n>FAST surpasses state-of-the-art baselines with superior debiasing performance.<n>This highlights the potential of fine-grained debiasing strategies to achieve fairness in large language models.
  arXiv  Detail & Related papers  (2024-08-07T17:14:58Z)
• The African Woman is Rhythmic and Soulful: An Investigation of Implicit Biases in LLM Open-ended Text Generation [3.9945212716333063]
  Implicit biases are significant because they influence the decisions made by Large Language Models (LLMs) Traditionally, explicit bias tests or embedding-based methods are employed to detect bias, but these approaches can overlook more nuanced, implicit forms of bias. We introduce two novel psychological-inspired methodologies to reveal and measure implicit biases through prompt-based and decision-making tasks.
  arXiv  Detail & Related papers  (2024-07-01T13:21:33Z)
• Bias-Conflict Sample Synthesis and Adversarial Removal Debias Strategy for Temporal Sentence Grounding in Video [67.24316233946381]
  Temporal Sentence Grounding in Video (TSGV) is troubled by dataset bias issue. We propose the bias-conflict sample synthesis and adversarial removal debias strategy (BSSARD)
  arXiv  Detail & Related papers  (2024-01-15T09:59:43Z)
• GPTBIAS: A Comprehensive Framework for Evaluating Bias in Large Language Models [83.30078426829627]
  Large language models (LLMs) have gained popularity and are being widely adopted by a large user community. The existing evaluation methods have many constraints, and their results exhibit a limited degree of interpretability. We propose a bias evaluation framework named GPTBIAS that leverages the high performance of LLMs to assess bias in models.
  arXiv  Detail & Related papers  (2023-12-11T12:02:14Z)
• General Greedy De-bias Learning [163.65789778416172]
  We propose a General Greedy De-bias learning framework (GGD), which greedily trains the biased models and the base model like gradient descent in functional space. GGD can learn a more robust base model under the settings of both task-specific biased models with prior knowledge and self-ensemble biased model without prior knowledge.
  arXiv  Detail & Related papers  (2021-12-20T14:47:32Z)
• Learning Debiased Models with Dynamic Gradient Alignment and Bias-conflicting Sample Mining [39.00256193731365]
  Deep neural networks notoriously suffer from dataset biases which are detrimental to model robustness, generalization and fairness. We propose a two-stage debiasing scheme to combat against the intractable unknown biases.
  arXiv  Detail & Related papers  (2021-11-25T14:50:10Z)
• Balancing out Bias: Achieving Fairness Through Training Reweighting [58.201275105195485]
  Bias in natural language processing arises from models learning characteristics of the author such as gender and race. Existing methods for mitigating and measuring bias do not directly account for correlations between author demographics and linguistic variables. This paper introduces a very simple but highly effective method for countering bias using instance reweighting.
  arXiv  Detail & Related papers  (2021-09-16T23:40:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.