CCCaption: Dual-Reward Reinforcement Learning for Complete and Correct Image Captioning

TL;DR

This paper introduces CCCaption, a reinforcement learning framework that optimizes image captions for completeness and correctness, addressing limitations of human-annotated references.

Contribution

It proposes a dual-reward reinforcement learning approach that explicitly enhances caption completeness and correctness using visual queries and hallucination penalties.

Findings

Consistent improvements across standard benchmarks.

Effective disentanglement of visual facts using diverse LVLMs.

Guides caption models beyond human-annotation imitation.

Abstract

Image captioning remains a fundamental task for vision language understanding, yet ground-truth supervision still relies predominantly on human-annotated references. Because human annotations reflect subjective preferences and expertise, ground-truth captions are often incomplete or even incorrect, which in turn limits caption models. We argue that caption quality should be assessed by two objective aspects: completeness (does the caption cover all salient visual facts?) and correctness (are the descriptions true with respect to the image?). To this end, we introduce CCCaption: a dual-reward reinforcement learning framework with a dedicated fine-tuning corpus that explicitly optimizes these properties to generate \textbf{C}omplete and \textbf{C}orrect \textbf{Captions}. For completeness, we use diverse LVLMs to disentangle the image into a set of visual queries, and reward captions that answer more of these queries, with a dynamic query sampling strategy to improve training efficiency. For correctness, we penalize captions that contain hallucinations by validating the authenticity of sub-caption queries, which are derived from the caption decomposition. Our symmetric dual-reward optimization jointly maximizes completeness and correctness, guiding models toward captions that better satisfy these objective criteria. Extensive experiments across standard captioning benchmarks show consistent improvements, offering a principled path to training caption models beyond human-annotation imitation.