NovGauge - Multi-Paper Grouping - Problem F1: leaderboard
Raw F1 for judging paper similarity on the problem dimension (technical challenges) in NovGauge multi-paper similarity grouping on 85 instances with full content (NovGauge, arXiv 2609.11234v1, Table 4). Percent, mean of three runs at the model's maximum reasoning setting; higher is better. Does not check evidence faithfulness; see Verified F1.
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Around February 2028. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Flash Lite | 80.2 | #200 |
| 2 | GPT-5.5 | 62.6 | #33 |
| 3 | Gemini 3.1 Pro (Preview) | 56.5 | #64 |
| 4 | Kimi K2.6 | 56.3 | #113 |
| 5 | Claude Sonnet 4.6 | 56 | #94 |
| 6 | Claude Opus 4.7 | 54.8 | #51 |
| 7 | GLM-5.1 | 52.2 | #109 |
| 8 | Qwen 3.6 Plus | 49.2 | #124 |
| 9 | MiniMax-M2.7 | 49 | #270 |
| 10 | GPT-5.4 Mini | 44.7 | #228 |
| 11 | DeepSeek V4 Flash | 43.3 | #155 |
| 12 | DeepSeek V4 Pro | 43 | #112 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=novgauge-multi-paper-grouping-problem-f1 · How It Works · Data refreshed daily, snapshot 2026-11-05.