UGC-VideoCaptioner: benchmark results
Provider: Other.
Aggregate Rating 1560 ± 39, rank #1552 of 10916 rated models, from 8 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CapRiCorn-1K | 11.8 | Keypoint accuracy (%; share of annotated keypoints a GPT-4.1 | 52.9 |
| CapRiCorn-1K - Referential Consistency | 3.6 | Subject referential consistency (%; share of same-subject de | 52.9 |
| video-SALMONN-2 Test Set (GPT-4.1 judge) - Total error | 48.6 | Missed + hallucinated events (%) | 50 |
| UGC-VideoCap - Average | 59.1 | Mean judge score (audio, visual, detail) | 41.7 |
| Daily-Omni (caption QA, Gemini-2.5-Pro reader) | 17 | Accuracy (%) | 40 |
| WorldSense (caption QA, Gemini-2.5-Pro reader) | 11.2 | Accuracy (%) | 40 |
| DiaDemBench - Speaker attribution F1 (REF) | 29.7 | REF F1 | 38.5 |
| DiaDemBench - Utterance transcription F1 (ASR) | 47 | ASR F1 | 38.5 |
Interactive version: theaggregate.ai/model?slug=ugc-videocaptioner · How It Works · Data refreshed daily, snapshot 2026-11-08.