AbstentionBench - false premise - KUQ/False assumptions - Recall: leaderboard
Metric: Abstention Recall (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 70.3 |
| 2 | Gemini 1.5 Pro | 67.67 |
| 3 | Llama-3.1-Tulu-3-70B | 66.17 |
| 4 | Llama-3.1-Tulu-3-70B-DPO | 65.79 |
| 5 | Llama 3.1 8B Instruct | 64.66 |
| 6 | Llama 3.1 405B Instruct | 62.78 |
| 7 | Qwen 2.5 32B Instruct | 62.41 |
| 8 | Llama 3.3 70B Instruct | 60.9 |
| 9 | Llama 3.1 70B Instruct | 57.52 |
| 10 | Llama-3.1-Tulu-3-8B | 56.77 |
| 11 | Llama-3.1-Tulu-3-8B-DPO | 55.64 |
| 12 | O1 | 54.89 |
| 13 | Mistral 7B Instruct (v0.3) | 51.5 |
| 14 | DeepSeek R1 Distill Llama 70B | 47.37 |
| 15 | Llama-3.1-Tulu-3-8B-SFT | 46.62 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-false-premise-kuq-false-assumptions-recall · How It Works · Data refreshed daily, snapshot 2026-09-19.