AI RESEARCH
ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison
arXiv CS.LG
•
ArXi:2605.20278v1 Announce Type: new Long-form image captioning exposes a reward granularity problem in RL: captions are judged as whole sequences, while the important errors occur at the level of individual visual claims. A good dense caption should be both faithful and informative, avoiding hallucination without omitting salient details. Yet pairwise preferences, reference-based metrics, and holistic scalar rewards compress these local errors into a single sequence-level signal, obscuring the tradeoff between factuality and coverage. We.