Jul 2026
Style over Substance: A Shortcut Audit of Emotion-Description Preference Evaluation
A systematic shortcut audit of EmoPrefer using content-blind probes shows that the current scores can be reached without verifying either description against the video, and recommends source-balanced pairing, strict length control, counter-stereotypical sliced reporting, and multi-annotator consensus for future cross-g...
Jia-Bing Yang, Yixiang Chen, Yuan Xu et al.
· arXiv.org · 0 citations