Salvato in:
| Autori principali: | , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2603.00207 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918361145802752 |
|---|---|
| author | Ghosal, Soumya Suvra Kim, Youngeun Li, Zhuowei Chaudhry, Ritwick Xu, Linghan Zhang, Hongjing Zablocki, Jakub Xing, Yifan Zhang, Qin |
| author_facet | Ghosal, Soumya Suvra Kim, Youngeun Li, Zhuowei Chaudhry, Ritwick Xu, Linghan Zhang, Hongjing Zablocki, Jakub Xing, Yifan Zhang, Qin |
| contents | Advances in large reasoning models have shown strong performance on complex reasoning tasks by scaling test-time compute through extended reasoning. However, recent studies observe that in vision-dependent tasks, extended textual reasoning at inference time can degrade performance as models progressively lose attention to visual tokens and increasingly rely on textual priors alone. To address this, prior works use reinforcement learning (RL)-based fine-tuning to route visual tokens or employ refocusing mechanisms during reasoning. While effective, these methods are computationally expensive, requiring large-scale data generation and policy optimization. To leverage the benefits of test-time compute without additional RL fine-tuning, we propose VisRef, a visually grounded test-time scaling framework. Our key idea is to actively guide the reasoning process by re-injecting a coreset of visual tokens that are semantically relevant to the reasoning context while remaining diverse and globally representative of the image, enabling more grounded multi-modal reasoning. Experiments on three visual reasoning benchmarks with state-of-the-art multi-modal large reasoning models demonstrate that, under fixed test-time compute budgets, VisRef consistently outperforms existing test-time scaling approaches by up to 6.4%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_00207 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | VisRef: Visual Refocusing while Thinking Improves Test-Time Scaling in Multi-Modal Large Reasoning Models Ghosal, Soumya Suvra Kim, Youngeun Li, Zhuowei Chaudhry, Ritwick Xu, Linghan Zhang, Hongjing Zablocki, Jakub Xing, Yifan Zhang, Qin Computer Vision and Pattern Recognition Artificial Intelligence Advances in large reasoning models have shown strong performance on complex reasoning tasks by scaling test-time compute through extended reasoning. However, recent studies observe that in vision-dependent tasks, extended textual reasoning at inference time can degrade performance as models progressively lose attention to visual tokens and increasingly rely on textual priors alone. To address this, prior works use reinforcement learning (RL)-based fine-tuning to route visual tokens or employ refocusing mechanisms during reasoning. While effective, these methods are computationally expensive, requiring large-scale data generation and policy optimization. To leverage the benefits of test-time compute without additional RL fine-tuning, we propose VisRef, a visually grounded test-time scaling framework. Our key idea is to actively guide the reasoning process by re-injecting a coreset of visual tokens that are semantically relevant to the reasoning context while remaining diverse and globally representative of the image, enabling more grounded multi-modal reasoning. Experiments on three visual reasoning benchmarks with state-of-the-art multi-modal large reasoning models demonstrate that, under fixed test-time compute budgets, VisRef consistently outperforms existing test-time scaling approaches by up to 6.4%. |
| title | VisRef: Visual Refocusing while Thinking Improves Test-Time Scaling in Multi-Modal Large Reasoning Models |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2603.00207 |