Mitigating Bias in Large Vision-Language Models via Counterfactual Ensemble Decoding

• URL: http://arxiv.org/abs/2608.21415v1
• Date: Wed, 12 Aug 2026 08:25:29 GMT
• Title: Mitigating Bias in Large Vision-Language Models via Counterfactual Ensemble Decoding
• Abstract summary: Large Vision-Language Models (LVLMs) have achieved remarkable performance across a wide range of tasks.<n>They often inherit social biases from their training data, resulting in biased behavior when processing portraits from different social groups.<n>We propose Counterfactual Ensemble Decoding (CED), a novel framework that constructs multi-group counterfactual perspectives.
• Score: 45.1503984011004
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Vision-Language Models (LVLMs) have achieved remarkable performance across a wide range of tasks; however, they often inherit social biases from their training data, resulting in biased behavior when processing portraits from different social groups. Existing debiasing approaches typically compare token probabilities between the original and biased generations during decoding, but they are fundamentally limited by their reliance on a single, stereotyped viewpoint and fail to account for the diversity of social perspectives. Inspired by the social science principle that diversity fosters fairness, we propose Counterfactual Ensemble Decoding (CED), a novel framework that constructs multi-group counterfactual perspectives within the visual representation space and integrates them during decoding to promote equitable model behavior. CED first performs counterfactual steering in the visual space by identifying semantic directions associated with each social group and generating counterfactual representations along these directions, thereby offering diverse perspectives that disrupt stereotypical narratives. During decoding, CED locates the decoder layer exhibiting the greatest divergence among these perspectives and ensembles their token distributions using uncertainty-aware weights, prioritizing high-confidence tokens from different groups to yield a more balanced probability distribution that guides fairer generation. Extensive experiments on three social bias evaluation benchmarks demonstrate that \tool achieves substantial improvements over leading baselines, reducing bias by up to 47.97% across scenarios involving occupations, descriptors, and persona traits. Moreover, CED also preserves the core capabilities of the original model with minimal degradation.

Related papers

• HoloFair: Unified T2I Fairness Evaluation and Fair-GRPO Debiasing [20.795940054868854]
  Existing evaluation methods address only single-dimensional biases.<n>We introduce HoloFair, a comprehensive benchmark framework for multidimensional demographic bias analysis.<n>We also introduce Fair-GRPO, a reinforcement-learning-based debiasing method.
  arXiv  Detail & Related papers  (2026-05-23T17:52:07Z)
• Interpretable Debiasing of Vision-Language Models for Social Fairness [55.85977929985967]
  We introduce an interpretable, model-agnostic bias mitigation framework, DeBiasLens, that localizes social attribute neurons in Vision-Language models.<n>We train SAEs on facial image or caption datasets without corresponding social attribute labels to uncover neurons highly responsive to specific demographics.<n>Our research lays the groundwork for future auditing tools, prioritizing social fairness in emerging real-world AI systems.
  arXiv  Detail & Related papers  (2026-02-27T13:37:11Z)
• Evaluating Social Bias in RAG Systems: When External Context Helps and Reasoning Hurts [7.344577590113121]
  Social biases inherent in large language models (LLMs) raise significant fairness concerns.<n>This work focuses on evaluating and understanding the social bias implications of RAG.
  arXiv  Detail & Related papers  (2026-02-10T06:27:56Z)
• Fair Deepfake Detectors Can Generalize [51.21167546843708]
  We show that controlling for confounders (data distribution and model capacity) enables improved generalization via fairness interventions.<n>Motivated by this insight, we propose Demographic Attribute-insensitive Intervention Detection (DAID), a plug-and-play framework composed of: i) Demographic-aware data rebalancing, which employs inverse-propensity weighting and subgroup-wise feature normalization to neutralize distributional biases; and ii) Demographic-agnostic feature aggregation, which uses a novel alignment loss to suppress sensitive-attribute signals.<n>DAID consistently achieves superior performance in both fairness and generalization compared to several state-of-the-art
  arXiv  Detail & Related papers  (2025-07-03T14:10:02Z)
• Perspectives in Play: A Multi-Perspective Approach for More Inclusive NLP Systems [3.011820285006942]
  This study proposes a new multi-perspective approach using soft labels to encourage the development of perspective aware models.<n>We conduct an analysis across diverse subjective text classification tasks, including hate speech, irony, abusive language, and stance detection.<n>Results show that the multi-perspective approach better approximates human label distributions, as measured by Jensen-Shannon Divergence (JSD)<n>Our approach exhibits lower confidence in tasks like irony and stance detection, likely due to the inherent subjectivity present in the texts.
  arXiv  Detail & Related papers  (2025-06-25T07:53:36Z)
• Biases Propagate in Encoder-based Vision-Language Models: A Systematic Analysis From Intrinsic Measures to Zero-shot Retrieval Outcomes [14.331322509462419]
  Social-group biases intrinsic to foundational encoder-based vision-language models (VLMs) manifest in biases in downstream tasks.<n>We introduce a controlled framework to measure this propagation by correlating intrinsic measures of bias in the representational space with measures of bias in zero-shot text-to-image (TTI) and image-to-text (ITT) retrieval.<n>Results show substantial correlations between intrinsic and extrinsic bias, with an average $rho$ = 0.83 $pm$ 0.10.<n> Notably, we find that larger/better-performing models exhibit greater bias propagation, a finding that raises concerns
  arXiv  Detail & Related papers  (2025-06-06T20:01:32Z)
• On the Fairness, Diversity and Reliability of Text-to-Image Generative Models [68.62012304574012]
  multimodal generative models have sparked critical discussions on their reliability, fairness and potential for misuse.<n>We propose an evaluation framework to assess model reliability by analyzing responses to global and local perturbations in the embedding space.<n>Our method lays the groundwork for detecting unreliable, bias-injected models and tracing the provenance of embedded biases.
  arXiv  Detail & Related papers  (2024-11-21T09:46:55Z)
• Social Bias Probing: Fairness Benchmarking for Language Models [38.180696489079985]
  This paper proposes a novel framework for probing language models for social biases by assessing disparate treatment. We curate SoFa, a large-scale benchmark designed to address the limitations of existing fairness collections. We show that biases within language models are more nuanced than acknowledged, indicating a broader scope of encoded biases than previously recognized.
  arXiv  Detail & Related papers  (2023-11-15T16:35:59Z)
• Regularizing Variational Autoencoder with Diversity and Uncertainty Awareness [61.827054365139645]
  Variational Autoencoder (VAE) approximates the posterior of latent variables based on amortized variational inference. We propose an alternative model, DU-VAE, for learning a more Diverse and less Uncertain latent space.
  arXiv  Detail & Related papers  (2021-10-24T07:58:13Z)
• Supercharging Imbalanced Data Learning With Energy-based Contrastive Representation Transfer [72.5190560787569]
  In computer vision, learning from long tailed datasets is a recurring theme, especially for natural image datasets. Our proposal posits a meta-distributional scenario, where the data generating mechanism is invariant across the label-conditional feature distributions. This allows us to leverage a causal data inflation procedure to enlarge the representation of minority classes.
  arXiv  Detail & Related papers  (2020-11-25T00:13:11Z)
• When Relation Networks meet GANs: Relation GANs with Triplet Loss [110.7572918636599]
  Training stability is still a lingering concern of generative adversarial networks (GANs) In this paper, we explore a relation network architecture for the discriminator and design a triplet loss which performs better generalization and stability. Experiments on benchmark datasets show that the proposed relation discriminator and new loss can provide significant improvement on variable vision tasks.
  arXiv  Detail & Related papers  (2020-02-24T11:35:28Z)

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.