AbstentionBench - underspecified context - MMLU Math - Precision: leaderboard
Metric: Abstention Precision (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.3 70B Instruct | 100 |
| 2 | Llama 3.1 405B Instruct | 100 |
| 3 | Gemini 1.5 Pro | 100 |
| 4 | O1 (High) | 100 |
| 5 | O1 (Low) | 100 |
| 6 | Llama 3.1 70B Instruct | 98.99 |
| 7 | O1 | 98.94 |
| 8 | Qwen 2.5 32B Instruct | 98.85 |
| 9 | Llama-3.1-Tulu-3-70B-DPO | 97.65 |
| 10 | Llama-3.1-Tulu-3-8B-SFT | 97.37 |
| 11 | GPT-4o | 96.94 |
| 12 | Llama-3.1-Tulu-3-8B | 96.77 |
| 13 | Llama 3.1 8B Instruct | 96.67 |
| 14 | Llama-3.1-Tulu-3-8B-DPO | 93.18 |
| 15 | DeepSeek R1 Distill Llama 70B | 91.84 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-mmlu-math-precision · How It Works · Data refreshed daily, snapshot 2026-09-19.