arXiv:2605.28023v1 Announce Type: new 
Abstract: Visual captioning requires models to capture visual content faithfully while minimizing both omission and hallucination. As the dominant paradigm for captioning, MLLMs have achieved strong performance through scaling and high-quality data. Recently, RL has emerged as a key route to driving MLLMs toward higher precision and broader coverage, however, existing reward designs for captioning fail to provide fine-grained and reliable signals for factual verification, limiting their effectiveness. To address this, we propose VCap, a Witness-Adjudicator reward that pairs the reference caption (a witness) with the visual signal (an adjudicator). By explicitly verifying factual consistency between the reference and policy-generated captions grounded in the visual signal, VCap delivers a reward signal with hypergeometric-distribution-level precision for caption quality verification. This design enables effective learning even from imperfect references, facilitating weak-to-strong generalization in RL training. In our experiments, an 8B model trained with VCap outperforms open- and closed-source SOTA models on multiple image and video captioning benchmarks. Human evaluation further confirms its strong alignment with factual correctness. Additionally, VCap improves MLLM perceptual capability, generalizes across tasks, and surpasses best-of-N distillation, challenging prior assumptions about RLVR.

تمثل المقدمة الحديثة لـ VCap، وهو نظام مكافآت Witness-Adjudicator، هدفًا لتحسين التسمية المرئية من خلال ضمان التناسق الواقعي بين التسميات المرجعية وتلك التي تم إنشاؤها بواسطة النماذج، مما يعزز جودة تمثيل المحتوى المرئي.

La reciente introducción de VCap, un sistema de recompensa Witness-Adjudicator, tiene como objetivo mejorar la subtitulación visual al garantizar la consistencia fáctica entre los subtítulos de referencia y los generados por los modelos, mejorando así la calidad de la representación del contenido visual.

La récente introduction de VCap, un système de récompense Witness-Adjudicator, vise à améliorer la légende visuelle en garantissant la cohérence factuelle entre les légendes de référence et celles générées par les modèles, améliorant ainsi la qualité de la représentation du contenu visuel.

The recent introduction of VCap, a Witness-Adjudicator reward system, aims to enhance visual captioning by ensuring factual consistency between reference captions and those generated by models, thus improving the quality of visual content representation.