Saved in:
Bibliographic Details
Main Authors: Du, Henghui, Zhou, Chang, Chen, Xi, Hu, Di
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.23823
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910038726017024
author Du, Henghui
Zhou, Chang
Chen, Xi
Hu, Di
author_facet Du, Henghui
Zhou, Chang
Chen, Xi
Hu, Di
contents Complex video reasoning, actually, relies excessively on fine-grained perception rather than on expert (e.g., Ph.D, Science)-level reasoning. Through extensive empirical observation, we have recognized the critical impact of perception. In particular, when perception ability is almost fixed, enhancing reasoning from Qwen3-8B to OpenAI-o3 yields only 0.7% performance improvement. Conversely, even minimal change in perception model scale (from 7B to 32B) boosts performance by 1.4%, indicating enhancing perception, rather than reasoning, is more critical to improve performance. Therefore, exploring how to enhance perception ability through reasoning without the need for expensive fine-grained annotation information is worthwhile. To achieve this goal, we specially propose APPO, the Attention-guided Perception Policy Optimization algorithm that leverages token-level dense rewards to improve model's fine-grained perception. The core idea behind APPO is to optimize those tokens from different responses that primarily focus on the same crucial video frame (called intra-group perception tokens). Experimental results on diverse video benchmarks and models with different scales (3/7B) demonstrate APPO consistently outperforms GRPO and DAPO (0.5%~4%). We hope our work provides a promising approach to effectively enhance model's perception abilities through reasoning in a low-cost manner, serving diverse scenarios and demands.
format Preprint
id arxiv_https___arxiv_org_abs_2602_23823
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle APPO: Attention-guided Perception Policy Optimization for Video Reasoning
Du, Henghui
Zhou, Chang
Chen, Xi
Hu, Di
Computer Vision and Pattern Recognition
Complex video reasoning, actually, relies excessively on fine-grained perception rather than on expert (e.g., Ph.D, Science)-level reasoning. Through extensive empirical observation, we have recognized the critical impact of perception. In particular, when perception ability is almost fixed, enhancing reasoning from Qwen3-8B to OpenAI-o3 yields only 0.7% performance improvement. Conversely, even minimal change in perception model scale (from 7B to 32B) boosts performance by 1.4%, indicating enhancing perception, rather than reasoning, is more critical to improve performance. Therefore, exploring how to enhance perception ability through reasoning without the need for expensive fine-grained annotation information is worthwhile. To achieve this goal, we specially propose APPO, the Attention-guided Perception Policy Optimization algorithm that leverages token-level dense rewards to improve model's fine-grained perception. The core idea behind APPO is to optimize those tokens from different responses that primarily focus on the same crucial video frame (called intra-group perception tokens). Experimental results on diverse video benchmarks and models with different scales (3/7B) demonstrate APPO consistently outperforms GRPO and DAPO (0.5%~4%). We hope our work provides a promising approach to effectively enhance model's perception abilities through reasoning in a low-cost manner, serving diverse scenarios and demands.
title APPO: Attention-guided Perception Policy Optimization for Video Reasoning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.23823