ProxyThinker: Test-Time Guidance through Small Visual Reasoners
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_ | 1866911180621086720 |
|---|---|
| author | Xiao, Zilin Koo, Jaywon Ouyang, Siru Hernandez, Jefferson Meng, Yu Ordonez, Vicente |
| author_facet | Xiao, Zilin Koo, Jaywon Ouyang, Siru Hernandez, Jefferson Meng, Yu Ordonez, Vicente |
| contents | Recent advancements in reinforcement learning with verifiable rewards have pushed the boundaries of the visual reasoning capabilities in large vision-language models (LVLMs). However, training LVLMs with reinforcement fine-tuning (RFT) is computationally expensive, posing a significant challenge to scaling model size. In this work, we propose ProxyThinker, an inference-time technique that enables large models to inherit the visual reasoning capabilities from small, slow-thinking visual reasoners without any training. By subtracting the output distributions of base models from those of RFT reasoners, ProxyThinker modifies the decoding dynamics and successfully elicits the slow-thinking reasoning demonstrated by the emerged sophisticated behaviors such as self-verification and self-correction. ProxyThinker consistently boosts performance on challenging visual benchmarks on spatial, mathematical, and multi-disciplinary reasoning, enabling untuned base models to compete with the performance of their full-scale RFT counterparts. Furthermore, our implementation efficiently coordinates multiple language models with parallelism techniques and achieves up to 38 $\times$ faster inference compared to previous decoding-time methods, paving the way for the practical deployment of ProxyThinker. Code is available at https://github.com/MrZilinXiao/ProxyThinker. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_24872 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | ProxyThinker: Test-Time Guidance through Small Visual Reasoners Xiao, Zilin Koo, Jaywon Ouyang, Siru Hernandez, Jefferson Meng, Yu Ordonez, Vicente Computer Vision and Pattern Recognition Artificial Intelligence Computation and Language Recent advancements in reinforcement learning with verifiable rewards have pushed the boundaries of the visual reasoning capabilities in large vision-language models (LVLMs). However, training LVLMs with reinforcement fine-tuning (RFT) is computationally expensive, posing a significant challenge to scaling model size. In this work, we propose ProxyThinker, an inference-time technique that enables large models to inherit the visual reasoning capabilities from small, slow-thinking visual reasoners without any training. By subtracting the output distributions of base models from those of RFT reasoners, ProxyThinker modifies the decoding dynamics and successfully elicits the slow-thinking reasoning demonstrated by the emerged sophisticated behaviors such as self-verification and self-correction. ProxyThinker consistently boosts performance on challenging visual benchmarks on spatial, mathematical, and multi-disciplinary reasoning, enabling untuned base models to compete with the performance of their full-scale RFT counterparts. Furthermore, our implementation efficiently coordinates multiple language models with parallelism techniques and achieves up to 38 $\times$ faster inference compared to previous decoding-time methods, paving the way for the practical deployment of ProxyThinker. Code is available at https://github.com/MrZilinXiao/ProxyThinker. |
| title | ProxyThinker: Test-Time Guidance through Small Visual Reasoners |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2505.24872 |