AffectSim (Active Observation): leaderboard
Metric: Macro-F1 (%; A-Obs: the view reached by the authors' two-stage active-observation baseline (person search, then target following), falling back to P-Init when no new view is acquired; five-class emotion recognition (angry, fearful, happy, neutral, sad) on the 3,941 test episodes of AffectSim, full-body affective motions replayed in Habitat indoor scenes; frozen recognizer, eight uniformly sampled frames for most open models, full video or eight frames for closed models). Source: arxiv.org. Saturation forecast: Around 2030. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 28.37 |
| 2 | Gemini 3 Pro | 25.42 |
| 3 | Gemini 3 Flash | 24.62 |
| 4 | GPT-5.6 Sol | 21.87 |
| 5 | GPT-5.2 | 21 |
| 6 | InternVL2.5-78B | 18.52 |
| 7 | Qwen 3 VL 8B Instruct | 13.19 |
| 8 | Qwen 3 VL 8B (Thinking) | 9.59 |
| 9 | Qwen2.5-Omni-7B | 9.45 |
| 10 | Qwen 2.5 VL 72B Instruct | 7.63 |
| 11 | MiniCPM-V-2.6 | 6.79 |
| 12 | Qwen 2.5 VL 7B Instruct | 6.11 |
Interactive version: theaggregate.ai/benchmark?slug=affectsim-active-observation · How It Works · Data refreshed daily, snapshot 2026-09-29.