跳到正文
arXiv cs.AI· Trung Nguyen Quang, Yuhao Dong, Shuo Sun, Shuai Liu, Shulin Tian, Kim-Hui Yap, Ziwei Liu·· 4 小时前AI 评分36

Video-HopChain:面向视频推理模型的多跳问题数据集与置信门控探索方法

Video-HopChain: Multi-Hop Questions and Confidence-Gated Exploration for Video Reasoning Models

AI 导读

研究者构建了 Video-HopChain,包含 13,378 个视频上的 22,550 道多跳问题及 1,000 题留出基准,每题串联 3 至 6 个 yes/no 子问题并以整数求和作为可验证奖励。

正文 · 原文

View PDF HTML (experimental)

Abstract:HopChain has shown on still images that multi-hop data synthesis improves vision-language reasoning, because long chain-of-thought reasoning exposes errors that compound across steps, while most data used for reinforcement learning with verifiable rewards (RLVR) rarely demands a chain of visual evidence, so these weaknesses are likely to stay unexposed. We observe the same problem in video, where this framework has not yet been explored. We therefore build Video-HopChain, a dataset of 22,550 multi-hop video questions over 13,378 videos, together with a held-out benchmark of 1,000 questions. Each question chains three to six yes/no questions about moments in one video, and each yields one of two integers depending on its answer. The final answer is the sum of these integers, so an exact match on that sum gives the verifiable reward that RLVR needs. We first train Qwen3-VL-8B with GRPO on a standard video dataset, and a second stage on Video-HopChain then raises the mean over eight video understanding and reasoning benchmarks from 55.4 to 57.9 and improves every one of them. Training on such a dataset, however, exposes a known limitation of GRPO: its learning signal comes from the reward variance within a group, so hard questions whose rollouts are all incorrect and easy questions whose rollouts are all correct both leave the group with no gradient. To recover these groups at the same compute budget, we introduce Confidence-Gated Exploration (CGE). With 8 rollouts per question, CGE samples the first 4 as usual. If these 4 are either all correct or all incorrect, it samples the last 4 with the policy's most confident token masked inside the reasoning span, and removes the masked positions from the loss while all 8 rollouts enter the advantage. With CGE, the mean rises further to 59.3. We release the dataset, the checkpoint, and the data generation and training code.
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as: arXiv:2609.25773 [cs.CV]
  (or arXiv:2609.25773v3 [cs.CV] for this version)
  https://doi.org/10.48550/arXiv.2609.25773

arXiv-issued DOI via DataCite

Submission history

From: Trung Nguyen Quang [view email]
[v1] Tue, 22 Sep 2026 07:08:13 UTC (442 KB)
[v2] Wed, 14 Oct 2026 12:32:11 UTC (1,053 KB)
[v3] Tue, 17 Nov 2026 14:33:04 UTC (1,736 KB)

来源:arXiv cs.AI · arxiv.org