AbstentionBench - underspecified context - Musique - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.3 70B Instruct | 83.08 |
| 2 | Llama 3.1 405B Instruct | 82.73 |
| 3 | Llama 3.1 8B Instruct | 82.38 |
| 4 | Gemini 1.5 Pro | 82.23 |
| 5 | Qwen 2.5 32B Instruct | 81.51 |
| 6 | Llama 3.1 70B Instruct | 80.9 |
| 7 | Mistral 7B Instruct (v0.3) | 77.45 |
| 8 | GPT-4o | 74.52 |
| 9 | Llama 3.1 8B | 70.55 |
| 10 | Llama 3.1 70B | 67.87 |
| 11 | Llama-3.1-Tulu-3-70B-DPO | 67.42 |
| 12 | O1 | 63.39 |
| 13 | Llama-3.1-Tulu-3-70B | 61.67 |
| 14 | DeepSeek R1 Distill Llama 70B | 51.46 |
| 15 | Llama-3.1-Tulu-3-8B-DPO | 31.04 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-musique-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.