AbstentionBench - underspecified context - ALCUNA - Recall: leaderboard
Metric: Abstention Recall (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 32B Instruct | 84.06 |
| 2 | Llama 3.1 8B Instruct | 78.23 |
| 3 | Llama-3.1-Tulu-3-70B-DPO | 77.02 |
| 4 | Llama-3.1-Tulu-3-70B | 76.96 |
| 5 | Gemini 1.5 Pro | 74.6 |
| 6 | GPT-4o | 74.54 |
| 7 | Llama 3.1 70B | 71.88 |
| 8 | Llama 3.3 70B Instruct | 70.5 |
| 9 | Llama 3.1 405B Instruct | 70.44 |
| 10 | Llama 3.1 70B Instruct | 70.27 |
| 11 | Mistral 7B Instruct (v0.3) | 67.84 |
| 12 | DeepSeek R1 Distill Llama 70B | 62.93 |
| 13 | Llama 3.1 8B | 53.94 |
| 14 | O1 | 53.06 |
| 15 | Llama-3.1-Tulu-3-8B-DPO | 52.42 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-alcuna-recall · How It Works · Data refreshed daily, snapshot 2026-09-19.