NormViz-Bench - Macro F1: leaderboard
NormViz-Bench macro F1 over the follow/avoid labels. Higher is better. Source: arxiv.org/abs/2609.06831.
Metric: Macro F1 (%). Source: arxiv.org. Saturation forecast: Around 2032. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 55.8 | #103 |
| 2 | GPT-5.2 | 45.6 | #115 |
| 3 | Qwen 2.5 3B | 44.6 | #1177 |
| 4 | Qwen 2.5 7B | 44.2 | #895 |
| 5 | Qwen 3 8B | 43.6 | #678 |
| 6 | GPT-4o | 41.5 | #357 |
| 7 | Qwen 3 4B | 37.7 | #834 |
| 8 | Gemma 3 4B | 34.6 | #1088 |
| 9 | Molmo2-8B | 31.5 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=normviz-bench-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-11-05.