GroupToM-Bench - Structural Constraint: leaderboard
Metric: mean judge score (0-100) on Level 5 structural constraint reasoning (how hierarchy, communication topology and norms distort behaviour), open-ended, scored 0 to 100 by a GPT-5 judge against expert references; 240 multimodal multi-agent scenarios (scene image plus dialogue); higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 59.7 |
| 2 | GPT-5 | 56.9 |
| 3 | GPT-5 Mini | 52.9 |
| 4 | GPT-4o | 47.2 |
| 5 | Claude Haiku 4.5 | 46.7 |
| 6 | InternVL3.5-8B | 41.4 |
| 7 | GPT-5 Nano | 41.1 |
| 8 | Qwen 2.5 VL 7B Instruct | 36.6 |
| 9 | Qwen 2 VL 7B | 35.1 |
Interactive version: theaggregate.ai/benchmark?slug=grouptom-bench-structural-constraint · How It Works · Data refreshed daily, snapshot 2026-09-29.