The authors replace free-form caption evaluation with three linked tracks.
Structured caption tests expose weaknesses in long-range audiovisual reasoning
OmniCapBench decomposes video descriptions into verifiable entities, visual shots and audio events, allowing errors to be localized rather than hidden inside one overall score.
Chinese Tech
Zhongyu Yang · Jiale Tao · Ruitao Chen · Zuhao Yang · Yingfang Yuan · Xueliang Zhao · +6 more
Hunyuan, Tencent · Nanyang Technological University · Northumbria University
Research Digest··3 min read
Yang and colleagues present a benchmark for diagnosing how multimodal language models describe continuous audio-visual content.
Why this paper
From Hunyuan, Tencent and 2 others
In one line
Audio-visual caption evaluation should use verifiable atomic units instead of free-form text for reliable, fine-grained diagnosis.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§