AbstentionBench - false premise - FalseQA - Recall: leaderboard
Metric: Abstention Recall (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro | 69 |
| 2 | O1 | 65.94 |
| 3 | Qwen 2.5 32B Instruct | 64.05 |
| 4 | Llama 3.3 70B Instruct | 60.7 |
| 5 | GPT-4o | 60.26 |
| 6 | Llama 3.1 405B Instruct | 56.62 |
| 7 | Llama 3.1 70B Instruct | 53.57 |
| 8 | Llama-3.1-Tulu-3-70B-DPO | 51.67 |
| 9 | Mistral 7B Instruct (v0.3) | 51.53 |
| 10 | Llama-3.1-Tulu-3-70B | 51.24 |
| 11 | Llama 3.1 8B Instruct | 47.74 |
| 12 | Llama-3.1-Tulu-3-8B-SFT | 43.23 |
| 13 | DeepSeek R1 Distill Llama 70B | 42.79 |
| 14 | Llama-3.1-Tulu-3-8B-DPO | 41.34 |
| 15 | Llama-3.1-Tulu-3-8B | 39.88 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-false-premise-falseqa-recall · How It Works · Data refreshed daily, snapshot 2026-09-19.