AbstentionBench - underspecified context - ALCUNA - Precision: leaderboard
Metric: Abstention Precision (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.1 405B Instruct | 93.27 |
| 2 | Mistral 7B Instruct (v0.3) | 92.89 |
| 3 | Llama 3.3 70B Instruct | 92.22 |
| 4 | DeepSeek R1 Distill Llama 70B | 91.67 |
| 5 | Llama 3.1 70B Instruct | 91.3 |
| 6 | Gemini 1.5 Pro | 89.54 |
| 7 | Llama-3.1-Tulu-3-8B-SFT | 89.29 |
| 8 | Llama-3.1-Tulu-3-70B-DPO | 87.36 |
| 9 | Llama-3.1-Tulu-3-70B | 87.18 |
| 10 | Llama-3.1-Tulu-3-8B | 87.03 |
| 11 | O1 | 85.97 |
| 12 | GPT-4o | 85.5 |
| 13 | Llama 3.1 8B Instruct | 84.06 |
| 14 | Llama-3.1-Tulu-3-8B-DPO | 83.46 |
| 15 | Qwen 2.5 32B Instruct | 82.45 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-alcuna-precision · How It Works · Data refreshed daily, snapshot 2026-09-19.