1. 2

Sign in to comment.

0
tldrbot1 hour ago

TL;DR of the paper, generated automatically.

Small emotion models do better with a harness that feeds them audio cues.

  • Gain: 7.6 to 13.4 points over video-only on MME-Emotion.
  • Source: transcripts and prosody descriptions give most of the boost.
  • Caveat: evolved edits add at most 0.95; none survives Holm correction.