Biases Propagate in Encoder-based Vision-Language Models: A Systematic Analysis From Intrinsic Measures to Zero-shot Retrieval Outcomes
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908397669974016 |
|---|---|
| author | Ghate, Kshitish Charlesworth, Tessa Diab, Mona Caliskan, Aylin |
| author_facet | Ghate, Kshitish Charlesworth, Tessa Diab, Mona Caliskan, Aylin |
| contents | To build fair AI systems we need to understand how social-group biases intrinsic to foundational encoder-based vision-language models (VLMs) manifest in biases in downstream tasks. In this study, we demonstrate that intrinsic biases in VLM representations systematically ``carry over'' or propagate into zero-shot retrieval tasks, revealing how deeply rooted biases shape a model's outputs. We introduce a controlled framework to measure this propagation by correlating (a) intrinsic measures of bias in the representational space with (b) extrinsic measures of bias in zero-shot text-to-image (TTI) and image-to-text (ITT) retrieval. Results show substantial correlations between intrinsic and extrinsic bias, with an average $ρ$ = 0.83 $\pm$ 0.10. This pattern is consistent across 114 analyses, both retrieval directions, six social groups, and three distinct VLMs. Notably, we find that larger/better-performing models exhibit greater bias propagation, a finding that raises concerns given the trend towards increasingly complex AI models. Our framework introduces baseline evaluation tasks to measure the propagation of group and valence signals. Investigations reveal that underrepresented groups experience less robust propagation, further skewing their model-related outcomes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_06506 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Biases Propagate in Encoder-based Vision-Language Models: A Systematic Analysis From Intrinsic Measures to Zero-shot Retrieval Outcomes Ghate, Kshitish Charlesworth, Tessa Diab, Mona Caliskan, Aylin Computation and Language To build fair AI systems we need to understand how social-group biases intrinsic to foundational encoder-based vision-language models (VLMs) manifest in biases in downstream tasks. In this study, we demonstrate that intrinsic biases in VLM representations systematically ``carry over'' or propagate into zero-shot retrieval tasks, revealing how deeply rooted biases shape a model's outputs. We introduce a controlled framework to measure this propagation by correlating (a) intrinsic measures of bias in the representational space with (b) extrinsic measures of bias in zero-shot text-to-image (TTI) and image-to-text (ITT) retrieval. Results show substantial correlations between intrinsic and extrinsic bias, with an average $ρ$ = 0.83 $\pm$ 0.10. This pattern is consistent across 114 analyses, both retrieval directions, six social groups, and three distinct VLMs. Notably, we find that larger/better-performing models exhibit greater bias propagation, a finding that raises concerns given the trend towards increasingly complex AI models. Our framework introduces baseline evaluation tasks to measure the propagation of group and valence signals. Investigations reveal that underrepresented groups experience less robust propagation, further skewing their model-related outcomes. |
| title | Biases Propagate in Encoder-based Vision-Language Models: A Systematic Analysis From Intrinsic Measures to Zero-shot Retrieval Outcomes |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2506.06506 |