Investigating Bias Representations in Llama 2 Chat via Activation Steering

• URL: http://arxiv.org/abs/2402.00402v1
• Date: Thu, 1 Feb 2024 07:48:50 GMT
• Title: Investigating Bias Representations in Llama 2 Chat via Activation Steering
• Authors: Dawn Lu, Nina Rimsky
• Abstract summary: We use activation steering to probe for and mitigate biases related to gender, race, and religion. Our findings reveal inherent gender bias in Llama 2 7B Chat, persisting even after Reinforcement Learning from Human Feedback. This work also provides valuable insights into effective red-teaming strategies for Large Language Models.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We address the challenge of societal bias in Large Language Models (LLMs), focusing on the Llama 2 7B Chat model. As LLMs are increasingly integrated into decision-making processes with substantial societal impact, it becomes imperative to ensure these models do not reinforce existing biases. Our approach employs activation steering to probe for and mitigate biases related to gender, race, and religion. This method manipulates model activations to direct responses towards or away from biased outputs, utilizing steering vectors derived from the StereoSet dataset and custom GPT4 generated gender bias prompts. Our findings reveal inherent gender bias in Llama 2 7B Chat, persisting even after Reinforcement Learning from Human Feedback (RLHF). We also observe a predictable negative correlation between bias and the model's tendency to refuse responses. Significantly, our study uncovers that RLHF tends to increase the similarity in the model's representation of different forms of societal biases, which raises questions about the model's nuanced understanding of different forms of bias. This work also provides valuable insights into effective red-teaming strategies for LLMs using activation steering, particularly emphasizing the importance of integrating a refusal vector.

Related papers

• The LLM Wears Prada: Analysing Gender Bias and Stereotypes through Online Shopping Data [8.26034886618475]
  We investigate whether Large Language Models can predict an individual's gender based solely on online shopping histories. Using a dataset of historical online purchases from users in the United States, we evaluate the ability of six LLMs to classify gender. Results indicate that while models can infer gender with moderate accuracy, their decisions are often rooted in stereotypical associations between product categories and gender.
  arXiv  Detail & Related papers  (2025-04-02T17:56:08Z)
• BiasConnect: Investigating Bias Interactions in Text-to-Image Models [73.76853483463836]
  We introduce BiasConnect, a novel tool designed to analyze and quantify bias interactions in Text-to-Image models. Our method provides empirical estimates that indicate how other bias dimensions shift toward or away from an ideal distribution when a given bias is modified. We demonstrate the utility of BiasConnect for selecting optimal bias mitigation axes, comparing different TTI models on the dependencies they learn, and understanding the amplification of intersectional societal biases in TTI models.
  arXiv  Detail & Related papers  (2025-03-12T19:01:41Z)
• How far can bias go? -- Tracing bias from pretraining data to alignment [54.51310112013655]
  This study examines the correlation between gender-occupation bias in pre-training data and their manifestation in LLMs. Our findings reveal that biases present in pre-training data are amplified in model outputs.
  arXiv  Detail & Related papers  (2024-11-28T16:20:25Z)
• The Root Shapes the Fruit: On the Persistence of Gender-Exclusive Harms in Aligned Language Models [58.130894823145205]
  We center transgender, nonbinary, and other gender-diverse identities to investigate how alignment procedures interact with pre-existing gender-diverse bias. Our findings reveal that DPO-aligned models are particularly sensitive to supervised finetuning. We conclude with recommendations tailored to DPO and broader alignment practices.
  arXiv  Detail & Related papers  (2024-11-06T06:50:50Z)
• Revealing and Reducing Gender Biases in Vision and Language Assistants (VLAs) [82.57490175399693]
  We study gender bias in 22 popular image-to-text vision-language assistants (VLAs) Our results show that VLAs replicate human biases likely present in the data, such as real-world occupational imbalances. To eliminate the gender bias in these models, we find that finetuning-based debiasing methods achieve the best tradeoff between debiasing and retaining performance on downstream tasks.
  arXiv  Detail & Related papers  (2024-10-25T05:59:44Z)
• Towards Implicit Bias Detection and Mitigation in Multi-Agent LLM Interactions [25.809599403713506]
  Large Language Models (LLMs) are increasingly being employed in numerous studies to simulate societies and execute diverse social tasks. LLMs are susceptible to societal biases due to their exposure to human-generated data. This study investigates the presence of implicit gender biases in multi-agent LLM interactions and proposes two strategies to mitigate these biases.
  arXiv  Detail & Related papers  (2024-10-03T15:28:05Z)
• Covert Bias: The Severity of Social Views' Unalignment in Language Models Towards Implicit and Explicit Opinion [0.40964539027092917]
  We evaluate the severity of bias toward a view by using a biased model in edge cases of excessive bias scenarios. Our findings reveal a discrepancy in LLM performance in identifying implicit and explicit opinions, with a general tendency of bias toward explicit opinions of opposing stances. The direct, incautious responses of the unaligned models suggest a need for further refinement of decisiveness.
  arXiv  Detail & Related papers  (2024-08-15T15:23:00Z)
• Subtle Biases Need Subtler Measures: Dual Metrics for Evaluating Representative and Affinity Bias in Large Language Models [10.73340009530019]
  This study addresses two such biases within Large Language Models (LLMs): representative bias and affinity bias. We introduce two novel metrics to measure these biases: the Representative Bias Score (RBS) and the Affinity Bias Score (ABS) Our analysis uncovers marked representative biases in prominent LLMs, with a preference for identities associated with being white, straight, and men. Our investigation of affinity bias reveals distinctive evaluative patterns within each model, akin to bias fingerprints'
  arXiv  Detail & Related papers  (2024-05-23T13:35:34Z)
• Locating and Mitigating Gender Bias in Large Language Models [40.78150878350479]
  Large language models (LLM) are pre-trained on extensive corpora to learn facts and human cognition which contain human preferences. This process can inadvertently lead to these models acquiring biases and prevalent stereotypes in society. We propose the LSDM (Least Square Debias Method), a knowledge-editing based method for mitigating gender bias in occupational pronouns.
  arXiv  Detail & Related papers  (2024-03-21T13:57:43Z)
• Self-Debiasing Large Language Models: Zero-Shot Recognition and Reduction of Stereotypes [73.12947922129261]
  We leverage the zero-shot capabilities of large language models to reduce stereotyping. We show that self-debiasing can significantly reduce the degree of stereotyping across nine different social groups. We hope this work opens inquiry into other zero-shot techniques for bias mitigation.
  arXiv  Detail & Related papers  (2024-02-03T01:40:11Z)
• Exploring the Jungle of Bias: Political Bias Attribution in Language Models via Dependency Analysis [86.49858739347412]
  Large Language Models (LLMs) have sparked intense debate regarding the prevalence of bias in these models and its mitigation. We propose a prompt-based method for the extraction of confounding and mediating attributes which contribute to the decision process. We find that the observed disparate treatment can at least in part be attributed to confounding and mitigating attributes and model misalignment.
  arXiv  Detail & Related papers  (2023-11-15T00:02:25Z)
• Evaluating Bias and Fairness in Gender-Neutral Pretrained Vision-and-Language Models [23.65626682262062]
  We quantify bias amplification in pretraining and after fine-tuning on three families of vision-and-language models. Overall, we find that bias amplification in pretraining and after fine-tuning are independent.
  arXiv  Detail & Related papers  (2023-10-26T16:19:19Z)
• 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)
• Learning Bias-Invariant Representation by Cross-Sample Mutual Information Minimization [77.8735802150511]
  We propose a cross-sample adversarial debiasing (CSAD) method to remove the bias information misused by the target task. The correlation measurement plays a critical role in adversarial debiasing and is conducted by a cross-sample neural mutual information estimator. We conduct thorough experiments on publicly available datasets to validate the advantages of the proposed method over state-of-the-art approaches.
  arXiv  Detail & Related papers  (2021-08-11T21:17:02Z)

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.