NovGauge - Multi-Paper Grouping - Problem Verified F1: leaderboard
Verified F1 on the problem dimension (technical challenges) in NovGauge multi-paper similarity grouping on 85 instances with full content: true positives count only if the cited evidence exists verbatim and an LLM judge finds it logically supports the stated reason (NovGauge, arXiv 2609.11234v1, Table 4). Percent; higher is better.
Metric: Verified F1 (%). Source: arxiv.org. Saturation forecast: Not forecast. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 25.2 | #33 |
| 2 | GPT-5.4 Mini | 14.6 | #228 |
| 3 | Kimi K2.6 | 9.9 | #113 |
| 4 | Gemini 3.1 Flash Lite | 9 | #200 |
| 5 | Gemini 3.1 Pro (Preview) | 8.9 | #64 |
| 6 | DeepSeek V4 Pro | 7 | #112 |
| 7 | GLM-5.1 | 6.3 | #109 |
| 8 | Qwen 3.6 Plus | 5.4 | #124 |
| 9 | Claude Sonnet 4.6 | 4.3 | #94 |
| 10 | DeepSeek V4 Flash | 3.7 | #155 |
| 11 | MiniMax-M2.7 | 1.8 | #270 |
| 12 | Claude Opus 4.7 | 0 | #51 |
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-verified-f1 · How It Works · Data refreshed daily, snapshot 2026-11-05.