Challenge: Current methods to learn biases from the perspective of model components are limited by their complexity and performance.
Approach: They propose a framework that incorporates causal mediation analysis to measure and map the pathways of bias generation and propagation within vision-language and multimodal tasks.
Outcome: The proposed framework is applicable to a wide range of vision-language and multimodal tasks and reduces bias by 22.03% and 9.04% in the MSCOCO and PASCAL-SENTENCE datasets.

Similar Papers

Mixed Signals: Decoding VLMs’ Reasoning and Underlying Bias in Vision-Language Conflict (2025.findings-emnlp)

Copied to clipboard

Challenge: Vision-language models have demonstrated impressive performance by effectively integrating visual and textual information to solve complex tasks.
Approach: They build upon existing benchmarks to create five datasets containing mismatched image-text pairs and examine how they reason over visual and textual data .
Outcome: The proposed model reasoned over visual and textual data in real-world applications but not in the visual and visual descriptions.
Think Before You Act: A Two-Stage Framework for Mitigating Gender Bias Towards Vision-Language Tasks (2024.naacl-long)

Copied to clipboard

Challenge: Existing vision-language models focus on salient attributes but ignore contextualized nuances, resulting in gender bias.
Approach: They propose a task-agnostic generation framework to mitigate gender bias in vision-language models.
Outcome: The proposed framework can mitigate gender bias in vision-language models . it yields all-sided but gender-obfuscated narratives, which prevents concentration on localized image features, especially gender attributes.
Quantifying and Mitigating Unimodal Biases in Multimodal Large Language Models: A Causal Perspective (2024.findings-emnlp)

Copied to clipboard

Challenge: Recent advances in Large Language Models have facilitated the development of Multimodal LLMs.
Approach: They propose a causal framework to interpret unimodal biases in visual question answering problems and a framework to integrate information from different modalities and mitigate biase.
Outcome: The proposed framework analyzes visual question answering (VQA) problems to assess their impact on predictions.
Biases Propagate in Encoder-based Vision-Language Models: A Systematic Analysis From Intrinsic Measures to Zero-shot Retrieval Outcomes (2025.findings-acl)

Copied to clipboard

Challenge: Existing encoder-based vision-language models (VLMs) contain intrinsic biases that manifest in biased outputs.
Approach: They propose a framework to measure intrinsic bias propagation by correlating intrinsic bias with extrinsic bias in zero-shot text-to-image and image-totext retrieval.
Outcome: The proposed framework shows that larger/better-performing models exhibit greater bias propagation, raising concerns given the trend towards increasingly complex AI models.
A Unified Framework and Dataset for Assessing Societal Bias in Vision-Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing studies have highlighted the existence of social biases within large vision and language models.
Approach: They propose a framework for systematically evaluating gender, race, and age biases in vision-language models with respect to professions.
Outcome: The proposed framework covers all supported inference modes of the recent vision-language models, including image-to-text, text-to image, and image- to-image.
BiasDora: Exploring Hidden Biased Associations in Vision-Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing studies on social biases focus on a limited set of documented associations, such as gender-profession or race-crime.
Approach: They propose to examine hidden, implicit bias associations across 9 bias dimensions by probing VLMs to uncover hidden, unexamined associations.
Outcome: The proposed methods reveal that biases vary in negativity, toxicity, and extremity.
Identifying and Mitigating Annotation Bias in Natural Language Understanding using Causal Mediation Analysis (2024.findings-acl)

Copied to clipboard

Challenge: Current NLU models obtain state-of-the-art accuracy on in-distribution benchmarks, but they use annotation bias to make predictions, negatively affecting the models' generalizability.
Approach: They apply causal mediation analysis to gauge how much each component mediates annotation biases and use causal-grounded masking and gradient unlearning to mitigate bias.
Outcome: The proposed methods improve the model's robustness against annotation bias even after employing other training-time debiasing techniques.
A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learning (2022.aacl-main)

Copied to clipboard

Challenge: Large-scale, pretrained vision-language models are growing in popularity due to impressive performance on downstream tasks with minimal finetuning.
Approach: They propose to apply ranking metrics to image-text representations to investigate bias measures and debiasing methods to reduce various bias measures.
Outcome: The proposed model reduces bias measures with minimal degradation to image-text representations.
VIGNETTE: Socially Grounded Bias Evaluation for Vision-Language Models (2026.acl-long)

Copied to clipboard

Challenge: Existing studies on VLM bias focus on portrait-style images and gender-occupation associations . existing studies ignore broader and more complex social stereotypes and their implied harm .
Approach: They propose a large-scale VQA benchmark for evaluating bias in vision-language models . they use a question-answering framework that spans factuality, perception, stereotyping, and decision making .
Outcome: The proposed framework examines bias in vision-language models using 30M+ images . findings reveal subtle, multifaceted, and surprising stereotypical patterns .
Do Images Speak Louder than Words? Investigating the Effect of Textual Misinformation in VLMs (2026.eacl-long)

Copied to clipboard

Challenge: Existing studies have shown that Vision-Language Models have robust multimodal reasoning capabilities, but their robustness against textual misinformation remains under-explored.
Approach: They propose to use visual-question-answering (VQA) prompts to generate persuasive prompts that deliberately conflict with visual evidence to test their models.
Outcome: The proposed framework shows that models are vulnerable to misleading prompts, and show an average performance drop of over 48.2% after only one round of persuasive conversation.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations