AbstentionBench - underspecified intent - KUQ/Ambiguous - Recall: leaderboard
Metric: Abstention Recall (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 75.83 |
| 2 | Gemini 1.5 Pro | 75.5 |
| 3 | Qwen 2.5 32B Instruct | 69.87 |
| 4 | Llama-3.1-Tulu-3-70B-DPO | 69.54 |
| 5 | Llama-3.1-Tulu-3-70B | 68.54 |
| 6 | O1 | 66.23 |
| 7 | Llama 3.1 8B Instruct | 65.56 |
| 8 | Llama 3.3 70B Instruct | 65.56 |
| 9 | Llama 3.1 405B Instruct | 65.56 |
| 10 | Llama 3.1 70B Instruct | 63.25 |
| 11 | Mistral 7B Instruct (v0.3) | 62.91 |
| 12 | Llama-3.1-Tulu-3-8B-DPO | 62.25 |
| 13 | Llama-3.1-Tulu-3-8B | 60.6 |
| 14 | Llama-3.1-Tulu-3-8B-SFT | 54.64 |
| 15 | Llama 3.1 8B | 48.68 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-intent-kuq-ambiguous-recall · How It Works · Data refreshed daily, snapshot 2026-09-19.