AffectSim (Initial View): leaderboard

Metric: Macro-F1 (%; P-Init: the agent's initial camera pose, held fixed; 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

#ModelScore
1Claude Opus 525.46
2Gemini 3 Flash24.77
3Gemini 3 Pro22.66
4GPT-5.220.89
5GPT-5.6 Sol19.29
6InternVL2.5-78B17.24
7Qwen 3 VL 8B Instruct10.12
8Qwen 3 VL 8B (Thinking)7.93
9Qwen2.5-Omni-7B7.78
10MiniCPM-V-2.66.55
11Qwen 2.5 VL 72B Instruct6.53
12Qwen 2.5 VL 7B Instruct6.09

Interactive version: theaggregate.ai/benchmark?slug=affectsim-initial-view · How It Works · Data refreshed daily, snapshot 2026-09-29.