aiqbench analysis
This research introduces ReVuE, an on-policy distillation framework that explicitly models the causal links between visual agent actions, observed evidence, and subsequent reasoning steps. By diagnosing failures in evidence acquisition, reading, or grounding stages, the method provides targeted token-level supervision that improves accuracy across perception and mathematical tasks without relying solely on auxiliary image views. The approach demonstrates significant gains over existing baselines while reducing redundancy in tool calls and reasoning traces for large multimodal models.
On-Policy Visual Evidence Distillation
arXiv:2609.36838v1 Announce Type: cross Abstract: Visual agents solve problems by interleaving reasoning with image operations, and on-policy distillation (OPD) provides guidance from a strong teacher on student-generated interaction trajectories. However, image operations change the evidence available for subsequent reasoning, so local errors in evidence acquisition (Acquire), reading (Read), or answer grounding (Ground) can propagate through the trajectory and lead to incorrect answers. Existing multimodal OPD methods primarily construct or contrast auxiliary views of the original image to strengthen supervision, without explicitly modeling the connections between student actions, resulting observations, and subsequent reasoning. This limits their ability to provide corrections tailored to different failure stages. We introduce Reflection on Visual Evidence (ReVuE), an on-policy distillation method for visual agents. ReVuE compares multiple student-generated trajectories for the same query, summarizes the observed visual evidence, and diagnoses the first failure across the Acquire, Read, and Ground stages. The resulting reflections provide training-time context for the teacher. We group and reweight token-level distillation losses according to how strongly these reflections affect the teacher's predictions. This design translates trajectory-level evidence diagnosis into targeted token-level supervision, guiding students to improve their visual evidence acquisition and reasoning. Across 11 benchmarks spanning the Qwen2.5-VL and InternVL3.5 model families, ReVuE outperforms all evaluated OPD baselines in weighted-average scores for perception, mathematical reasoning, and general tasks. ReVuE also reduces redundancy in reasoning and tool calls while improving tool-call accuracy and task accuracy. Code is available at https://github.com/sylvain-wei/ReVuE
Abstract:Visual agents solve problems by interleaving reasoning with image operations, and on-policy distillation (OPD) provides guidance from a strong teacher on student-generated interaction trajectories. However, image operations change the evidence available for subsequent reasoning, so local errors in evidence acquisition (Acquire), reading (Read), or answer grounding (Ground) can propagate through the trajectory and lead to incorrect answers. Existing multimodal OPD methods primarily construct or contrast auxiliary views of the original image to strengthen supervision, without explicitly modeling the connections between student actions, resulting observations, and subsequent reasoning. This limits their ability to provide corrections tailored to different failure stages. We introduce Reflection on Visual Evidence (ReVuE), an on-policy distillation method for visual agents. ReVuE compares multiple student-generated trajectories for the same query, summarizes the observed visual evidence, and diagnoses the first failure across the Acquire, Read, and Ground stages. The resulting reflections provide training-time context for the teacher. We group and reweight token-level distillation losses according to how strongly these reflections affect the teacher's predictions. This design translates trajectory-level evidence diagnosis into targeted token-level supervision, guiding students to improve their visual evidence acquisition and reasoning. Across 11 benchmarks spanning the Qwen2.5-VL and InternVL3.5 model families, ReVuE outperforms all evaluated OPD baselines in weighted-average scores for perception, mathematical reasoning, and general tasks. ReVuE also reduces redundancy in reasoning and tool calls while improving tool-call accuracy and task accuracy. Code is available at this https URL
44 pages, including appendices. Project page: this https URL . Code: this https URL
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2609.36838 [cs.CV]
(or arXiv:2609.36838v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.36838 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 352 words
Related coverage
ViCuR: Visual Cues as Recoverable Privilege for Multimodal On-Policy Distillation
arXiv · Jun 6, 2026
Med-OPD: Improving Medical Vision-Language Models via Evidence-Aware On-Policy Distillation
arXiv · Jul 21, 2026
Evidence-Aligned Multimodal On-Policy Self-Distillation for Fine-Grained Visual Understanding
arXiv · Sep 29, 2026
VC-OPD: Visual Counterfactual On-Policy Distillation for Grounded Vision-Language Reasoning
arXiv · Oct 15, 2026