Tuna : Comprehensive Fine-grained Temporal Understanding Evaluation on Dense Dynamic Videos

Fanheng Kong, Jingyuan Zhang, Hongzhi Zhang, Shi Feng, Daling Wang, Linhao Yu, Xingguang Ji, Yu Tian, Victoria W, Fuzheng Zhang · 2025

Read Tuna : Comprehensive Fine-grained Temporal Understanding Evaluation on Dense Dynamic Videos by Fanheng Kong, Jingyuan Zhang, Hongzhi Zhang, Shi Feng, Daling Wang, Linhao Yu, Xingguang Ji, Yu Tian, Victoria W, Fuzheng Zhang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-20124.txt.

Description

Fanheng Kong and colleagues build a benchmark that grades video models on time itself: how actions, scenes, cameras, and objects change together. Existing video tests isolate one aspect or skim surfaces. Their suite uses dense, dynamic clips and poses two tasks, captioning and question answering. Writing in twenty twenty-five, the team designed interpretable, robust scoring criteria and varied the scenarios widely. Several leading models were then graded dimension by dimension, giving a fine-grained report card rather than a single number.

The weaknesses are specific. Models describe actions thinly. They lose track when several subjects act at once. They barely notice camera motion. Each failure mode points at a concrete fix for the next model generation. Data and code are public, so the benchmark can be rerun as models improve.

Catalog record

CREATORFanheng Kong, Jingyuan Zhang, Hongzhi Zhang, Shi Feng, Daling Wang, Linhao Yu, Xingguang Ji, Yu Tian, Victoria W, Fuzheng Zhang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-20124

Browse the full library · About the Tower of Babel