AbstentionBench - underspecified context - MMLU History - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro | 71.88 |
| 2 | GPT-4o | 67.8 |
| 3 | Qwen 2.5 32B Instruct | 62.3 |
| 4 | O1 | 60.71 |
| 5 | Llama-3.1-Tulu-3-70B-DPO | 58.18 |
| 6 | Llama 3.1 8B Instruct | 53.57 |
| 7 | Llama-3.1-Tulu-3-70B | 52.83 |
| 8 | O1 (High) | 52.83 |
| 9 | O1 (Low) | 52.83 |
| 10 | Llama 3.1 70B Instruct | 51.85 |
| 11 | Llama 3.1 405B Instruct | 47.27 |
| 12 | TinyLlama-1.1B-Chat-v1.0 | 34.62 |
| 13 | Llama 3.3 70B Instruct | 33.33 |
| 14 | Llama-3.1-Tulu-3-8B | 26.67 |
| 15 | Llama-3.1-Tulu-3-8B-DPO | 26.67 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-mmlu-history-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.