captioning

A user complains that streaming services like Amazon Prime are replacing accurate, human-made subtitles on older shows with error-ridden AI-generated subtitles, questioning why existing high-quality captions aren't reused.

OmniCapBench 提出一个深度结构化的音视频字幕评估框架，将预测目标从自由文本转为可验证的原子评估单元（实体引用、视觉镜头、音频事件），在 786 个密集标注视频上区分多模态大模型的感知错误，揭示前沿模型长时程音视觉推理能力薄弱。

Introduces OmniCap-IF, the first comprehensive benchmark for evaluating instruction-following in omni-modal video captioning, revealing a format-content tradeoff and proposing improved models and datasets.

This paper introduces Video2LoRA, a method that predicts Low-Rank Adaptation (LoRA) weights directly from video representations, enabling efficient video processing in frozen vision-language models. It reduces visual token load by up to 1500x and query TTFT by 6-80x while maintaining performance on video summarization and captioning benchmarks.