AVIS: Adaptive Test-Time Scaling for Vision-Language Models
About
Modern Vision-Language Models (VLMs) benefit from chain-of-thought prompting and test-time scaling, but these gains often come with prohibitive inference cost due to large visual contexts and long decoding chains. We view this cost through two coupled axes: Visual Context Scaling (VCS), which controls how much visual evidence is passed to the language model, and Visual Reasoning Scaling (VRS), which controls how much inference-time reasoning search is performed. Existing methods typically optimize one axis at a time, leaving the joint allocation of compute across these axes underexplored. We introduce Adaptive Visual Inference Scaling (AVIS), a lightweight policy that adapts both VCS and VRS per query. AVIS realizes VCS through Key Diversity Visual (KDV) pruning, a training-free $O(N)$ key-based rule for removing redundant visual tokens before prefilling, and realizes VRS through adaptive self-consistency, using a learned difficulty predictor to select the number of reasoning rollouts. AVIS is deployment-friendly and compatible with shared-prefill inference, where all rollouts reuse a single prefilling pass and KV cache. Across diverse image and video reasoning benchmarks, AVIS improves the accuracy--compute trade-off relative to VCS-only and VRS-only baselines, and remains effective on top of RL post-trained VLMs while keeping compute and latency low.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multimodal Evaluation | MME | Score2.47e+3 | 902 | |
| Multimodal Evaluation | MMStar | -- | 177 | |
| Object Hallucination Evaluation | POPE | POPE Score89.1 | 22 | |
| General VQA | MME | VQA Score2.37e+3 | 19 | |
| Mathematical Reasoning | MathVision | Score41.1 | 12 | |
| Mathematical Reasoning | MathVista | Score71.8 | 12 | |
| Visual Reasoning | CVBench | Score76.1 | 12 | |
| Video VQA | Video-TT | Score38.4 | 7 | |
| Video VQA | Q-Bench-Video | Score61.3 | 7 | |
| Image VQA | MMMU-Pro | Score43.95 | 7 |