SOURCE-LINKED INTELLIGENCE

Making Every Tool Call Count: Necessary Tool-Evidence Path Rewards for Agentic Vision-Language Models

arXiv · AI, language, vision and robotics · article · Sep 3, 2026 · UTC

Modern vision-language models (VLMs) can directly answer many image-grounded questions, yet they often struggle with complex queries requiring fine-grained visual details or external knowledge. To acquire this missing evidence, agentic VLMs invoke tools such as image cropping, image search, and text search. However, existing training paradigms primarily evaluate tool-use based on final answer correctness, leaving evidence acquisition and utilization insufficiently supervised. This leads to two critical shortcomings: (i) models frequently issue redundant or off-target tool calls that fail to ga

Read original source ↗ Open in workspace

recordType

paper

region

Global