EviRover improves image understanding by gathering extra visual evidence
EviRover: Reinforcing Agentic Perception Beyond a Glance
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Most visual AI systems try to understand a picture in one quick look, but that can miss important details or recent information. The authors created EviRover, a system that can examine images more carefully and interactively to find the evidence needed for tricky questions. They built special datasets and a benchmark to train and test this ability. EviRover outperforms previous models on detailed image understanding and can also generalize to other visual tasks.
What this means in practice
- •For computer vision teams: Improve image analysis systems to handle subtle details by enabling models to gather additional visual information beyond a single glance.
- •For multimodal ai developers: Build AI models that can interactively verify and refine visual information, resulting in better performance on updated or complex visual data.
Authors
Kaixuan Fan, Kaituo Feng, Tianshuo Peng, Yilei Jiang, Manyuan Zhang, Junke Wang, Xiangyu Yue
Abstract
Visual perception is conventionally formulated as a one-shot prediction from a single glance at the image, under the assumption that the image content and the model's parametric knowledge suffice to resolve the query. This assumption often fails in real-world scenarios that hinge on fine-grained visual details or require knowledge-intensive and up-to-date information. We term such cases \textit{perception under insufficient evidence} and formulate perception as an agentic process that can obtain information beyond a single glance. To address the absence of data for this setting, we design two dedicated data generation pipelines, yielding EviRover-SFT-5K and EviRover-RL-12K for training. We further construct EviLens, a human-verified benchmark comprising 688 instances across five perception categories. Building on these data, we present EviRover, to our knowledge the first perception agent explicitly trained to resolve perceptual queries through interaction, using supervised fine-tuning followed by agentic reinforcement learning. Experiments show that the 4B EviRover outperforms its backbone by 30 points on average on EviLens, reaching performance comparable to advanced proprietary models. The gains transfer beyond EviLens to WebEyes, conventional perception benchmarks, and general multimodal benchmarks, including a 15-point improvement on BrowseComp-VL. All code, models, and data are released.