Look Where It Matters: High-Resolution Crops Retrieval for Efficient VLMs
This addresses the problem of computational inefficiency in VLMs for applications requiring fine visual details, representing an incremental improvement through a novel hybrid method.
The paper tackles the accuracy-efficiency trade-off in vision-language models by introducing AwaRes, a framework that uses low-resolution global views and tool-calling to retrieve high-resolution segments as needed, achieving improved performance with reduced computational costs.
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