Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.11109 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917141026963456 |
|---|---|
| author | Ahmadpour, Mohammadjavad Meighani, Amirmahdi Taebi, Payam Ghahroodi, Omid Izadi, Amirmohammad Baghshah, Mahdieh Soleymani |
| author_facet | Ahmadpour, Mohammadjavad Meighani, Amirmahdi Taebi, Payam Ghahroodi, Omid Izadi, Amirmohammad Baghshah, Mahdieh Soleymani |
| contents | Test-time scaling (TTS) has emerged as a powerful paradigm for improving the reasoning ability of Large Language Models (LLMs) by allocating additional computation at inference, yet its application to multimodal systems such as Vision-Language Models (VLMs) remains underexplored. In this work, we present a systematic empirical study of inference time reasoning methods applied across both open-source and closed-source VLMs on different benchmarks. Our results reveal that while closed-source models consistently benefit from structured reasoning and iterative Self-Refinement, open-source VLMs show inconsistent behavior: external verification provides the most reliable gains, whereas iterative refinement often degrades performance. We further find that the effectiveness of TTS is dataset-dependent, yielding clear improvements on multi-step reasoning tasks but offering only limited gains on perception-focused benchmarks. These findings demonstrate that TTS is not a universal solution and must be tailored to both model capabilities and task characteristics, motivating future work on adaptive TTS strategies and multimodal reward models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_11109 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Limits and Gains of Test-Time Scaling in Vision-Language Reasoning Ahmadpour, Mohammadjavad Meighani, Amirmahdi Taebi, Payam Ghahroodi, Omid Izadi, Amirmohammad Baghshah, Mahdieh Soleymani Machine Learning Test-time scaling (TTS) has emerged as a powerful paradigm for improving the reasoning ability of Large Language Models (LLMs) by allocating additional computation at inference, yet its application to multimodal systems such as Vision-Language Models (VLMs) remains underexplored. In this work, we present a systematic empirical study of inference time reasoning methods applied across both open-source and closed-source VLMs on different benchmarks. Our results reveal that while closed-source models consistently benefit from structured reasoning and iterative Self-Refinement, open-source VLMs show inconsistent behavior: external verification provides the most reliable gains, whereas iterative refinement often degrades performance. We further find that the effectiveness of TTS is dataset-dependent, yielding clear improvements on multi-step reasoning tasks but offering only limited gains on perception-focused benchmarks. These findings demonstrate that TTS is not a universal solution and must be tailored to both model capabilities and task characteristics, motivating future work on adaptive TTS strategies and multimodal reward models. |
| title | Limits and Gains of Test-Time Scaling in Vision-Language Reasoning |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2512.11109 |