InternVideo2-Chat-8B: benchmark results
Provider: Other.
Aggregate Rating 1518 ± 48, rank #2093 of 10916 rated models, from 9 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MA-Bench (Closed-Ended) | 45.39 | Accuracy (%) averaged over the six closed-ended tasks (coars | 76.2 |
| MA-Bench - Coarse-Grained Recognition | 22.9 | Accuracy (%) on the coarse-grained (body-level) micro-action | 76.2 |
| MA-Bench - Fine-Grained Recognition | 28.1 | Accuracy (%) on the fine-grained (action-level) micro-action | 52.4 |
| EmoBench-M | 51.5 | Average Score | 38 |
| AffectSim (Active Observation) | 8.07 | Macro-F1 (%; A-Obs: the view reached by the authors' two-sta | 17.4 |
| AffectSim (Initial View) | 7.04 | Macro-F1 (%; P-Init: the agent's initial camera pose, held f | 17.4 |
| AffectSim (Monitor View) | 7.04 | Macro-F1 (%; P-Mon: a fixed room-level monitoring camera; fi | 17.4 |
| AffectSim (Reference View) | 9.98 | Macro-F1 (%; P-Ref: an informative camera trajectory chosen | 17.4 |
| MA-Bench (Open-Ended) | 0.06 | GPT-4o judge score (0 to 5) averaged over the six open-ended | 7.1 |
Interactive version: theaggregate.ai/model?slug=internvideo2-chat-8b · How It Works · Data refreshed daily, snapshot 2026-11-08.