BoxComm - Commentary Rhythm (KL): leaderboard
BoxComm streaming commentary rhythm measured as the KL divergence between the per-minute distributions of human and generated commentary time across the play-by-play, tactical and contextual categories. Lower is better, with 0 meaning the human pacing is matched exactly.
Metric: KL divergence (0 or more, unbounded) between the per-minute human and predicted duration distributions over the three commentary categories, averaged over minutes, in streaming commentary rhythm assessment on BoxComm-Eval: the model watches each match in short chunks and comments freely with a generic commentator prompt, GPT-5.2 classifies each generated sentence as play-by-play, tactical or contextual, and the timing is compared with the human broadcast commentary; zero-shot; lower is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | LiveCC | 0.037 | #1839 |
| 2 | StreamingVLM | 0.23 | #890 |
Interactive version: theaggregate.ai/benchmark?slug=boxcomm-commentary-rhythm-kl · How It Works · Data refreshed daily, snapshot 2026-11-21.