Claim-Level Rubric Rewards for Video Caption Reinforcement Learning
2026-07-06 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors present a new method called Claim-Level Rubric Rewards (CuRe) to improve how computers learn to describe videos accurately. Instead of judging entire captions at once or comparing them strictly to example captions, CuRe breaks down captions into smaller, detailed claims and checks each one carefully. This helps avoid problems like missing facts or awkward writing that can happen with previous methods. Their approach focuses on verifying specific parts of captions to make learning more reliable.
Reinforcement LearningDense Video CaptioningReward DesignClaim-Level VerificationRubricFactual AccuracyReference-Based EvaluationOpen-Ended GenerationReward Hacking
Authors
Mingqi Gao, Hongyuan Dong, Yifei Chen, Zhisheng Zhong, Zheng Ruan, Wenjin Hou, Yu Chen, Han Hu, Yansong Tang
Abstract
In this paper, we introduce Claim-Level Rubric Rewards (CuRe), a structured reward framework designed to address the reward-design bottleneck in reinforcement learning for dense video captioning. Existing reward designs generally fall into two categories: holistic response-level judgment across heterogeneous criteria, or alignment-based evaluation against reference captions. However, both paradigms suffer from fundamental limitations. Holistic rewards struggle to ensure factual accuracy and are prone to stylistic reward hacking, while reference-based rewards overly rely on rigid textual alignment, failing to preserve the completeness and diversity inherent to open-ended generation tasks. To address these challenges, CuRe reformulates reward modeling as fine-grained claim-level verification. Specifically, CuRe decomposes captions into category-aware atomic claims through a structured rubric, converting holistic evaluation into simpler and more reliable claim-level verification.