Look Where It Matters: High-Resolution Crops Retrieval for Efficient VLMs

arXiv:2603.16932 · cs.CV, cs.AI · Submitted 2026-03-14 · Read on arXiv

cs.CV, cs.AI

Submitted: 2026-03-14

Updated: 2026-09-23

Code: https://github.com/huggingface/trl

Project page: https://nimrodshabtay.github.io/AwaRes

License: http://creativecommons.org/licenses/by-sa/4.0/

The gist: Vision-language models (VLMs) typically process images at a native high-resolution, forcing a trade-off between accuracy and computational efficiency: high-resolution inputs capture fine details but

Terminology

Abstract

Vision-language models (VLMs) typically process images at a native high-resolution, forcing a trade-off between accuracy and computational efficiency: high-resolution inputs capture fine details but incur significant computational costs, while low-resolution inputs advocate for efficiency, they potentially miss critical visual information, like small text. We present AwaRes, a spatial-on-demand framework that resolves this accuracy-efficiency trade-off by operating on a low-resolution global view and using tool-calling to retrieve only high-resolution segments needed for a given query. We construct supervised data automatically: a judge compares low- vs. high-resolution answers to label whether cropping is needed, and an oracle grounding model localizes the evidence for the correct answer, which we map to a discrete crop set to form multi-turn tool-use trajectories. We train our framework with cold-start SFT followed by multi-turn GRPO with a composite reward that combines semantic answer correctness with explicit crop-cost penalties. Project page: https://nimrodshabtay.github.io/AwaRes

Sources

Related papers