AbstentionBench - underspecified context - MMLU Math - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.1 70B Instruct | 84.48 |
| 2 | O1 (Low) | 83.33 |
| 3 | GPT-4o | 82.25 |
| 4 | O1 | 81.94 |
| 5 | Llama 3.1 405B Instruct | 81.78 |
| 6 | O1 (High) | 81.25 |
| 7 | Qwen 2.5 32B Instruct | 78.18 |
| 8 | Llama 3.1 8B Instruct | 78.03 |
| 9 | Llama-3.1-Tulu-3-70B-DPO | 76.15 |
| 10 | Mistral 7B Instruct (v0.3) | 71 |
| 11 | Gemini 1.5 Pro | 67.66 |
| 12 | Llama 3.3 70B Instruct | 64.97 |
| 13 | DeepSeek R1 Distill Llama 70B | 49.45 |
| 14 | Llama-3.1-Tulu-3-8B-DPO | 46.33 |
| 15 | Llama-3.1-Tulu-3-8B-SFT | 43.27 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-mmlu-math-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.