VisualDeltas: Learning Preferences from Visual Quality Perturbations
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914378336436224 |
|---|---|
| author | Huang, Hailiang Liu, Yihao Guan, Shengyue Li, Haoze Li, Sujian |
| author_facet | Huang, Hailiang Liu, Yihao Guan, Shengyue Li, Haoze Li, Sujian |
| contents | We present VisualDeltas, a lightweight preference-learning framework that extracts supervision from visual quality variations in multimodal data. By leveraging the systematic impact of image quality on visual perception and reasoning, VisualDeltas induces informative preference signals without relying on human annotations or external teachers. The framework supports both label-free and label-based regimes, enabling flexible use of available supervision when present. Across diverse multimodal benchmarks and model scales, VisualDeltas consistently outperforms rejection-sampling fine-tuning and improves generalization, and extends naturally to a range of visual degradations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_07272 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | VisualDeltas: Learning Preferences from Visual Quality Perturbations Huang, Hailiang Liu, Yihao Guan, Shengyue Li, Haoze Li, Sujian Artificial Intelligence We present VisualDeltas, a lightweight preference-learning framework that extracts supervision from visual quality variations in multimodal data. By leveraging the systematic impact of image quality on visual perception and reasoning, VisualDeltas induces informative preference signals without relying on human annotations or external teachers. The framework supports both label-free and label-based regimes, enabling flexible use of available supervision when present. Across diverse multimodal benchmarks and model scales, VisualDeltas consistently outperforms rejection-sampling fine-tuning and improves generalization, and extends naturally to a range of visual degradations. |
| title | VisualDeltas: Learning Preferences from Visual Quality Perturbations |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2603.07272 |