NormGym - Task Success (Seen Layout): leaderboard

Interactive social-norm learning benchmark (arXiv:2611.08613): task-goal completion score (%) over 150 seen-layout episodes across 50 hidden norms (1.0 for two goals, 0.5 for one, 0 for none).

Metric: Task Goal Completion Score (%), Seen Layout. Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked. Updated 20 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 595.7#26
2Gemini 3.1 Pro (Preview)90.3#101
3GPT-5.6 Sol86.3#36
4Ministral-3-8B-Instruct49.7#1299
5Qwen 3 VL 8B Instruct38.3#1035
6GLM-4.6V-Flash11.7#1047
7Llama 3.2 11B Instruct3#5189

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=normgym-task-success-seen-layout · How It Works · Data refreshed daily, snapshot 2026-11-20.