AbstentionBench - underspecified context - MMLU Math - Recall: leaderboard
Metric: Abstention Recall (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.1 70B Instruct | 73.68 |
| 2 | GPT-4o | 71.43 |
| 3 | O1 (Low) | 71.43 |
| 4 | O1 | 69.92 |
| 5 | Llama 3.1 405B Instruct | 69.17 |
| 6 | O1 (High) | 68.42 |
| 7 | Llama 3.1 8B Instruct | 65.41 |
| 8 | Qwen 2.5 32B Instruct | 64.66 |
| 9 | Llama-3.1-Tulu-3-70B-DPO | 62.41 |
| 10 | Mistral 7B Instruct (v0.3) | 61.65 |
| 11 | Gemini 1.5 Pro | 51.13 |
| 12 | Llama 3.3 70B Instruct | 48.12 |
| 13 | DeepSeek R1 Distill Llama 70B | 33.83 |
| 14 | Llama-3.1-Tulu-3-8B-DPO | 30.83 |
| 15 | Llama-3.1-Tulu-3-8B-SFT | 27.82 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-mmlu-math-recall · How It Works · Data refreshed daily, snapshot 2026-09-19.