AbstentionBench - false premise - KUQ/False assumptions - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 79.41 |
| 2 | Gemini 1.5 Pro | 78.09 |
| 3 | Llama-3.1-Tulu-3-70B | 76.36 |
| 4 | Llama-3.1-Tulu-3-70B-DPO | 76.09 |
| 5 | Llama 3.3 70B Instruct | 72.81 |
| 6 | Qwen 2.5 32B Instruct | 72.65 |
| 7 | Llama 3.1 405B Instruct | 71.67 |
| 8 | Llama 3.1 8B Instruct | 70.64 |
| 9 | Llama-3.1-Tulu-3-8B | 70.07 |
| 10 | O1 | 69.52 |
| 11 | Llama-3.1-Tulu-3-8B-DPO | 69.32 |
| 12 | Llama 3.1 70B Instruct | 69.07 |
| 13 | Mistral 7B Instruct (v0.3) | 64.93 |
| 14 | DeepSeek R1 Distill Llama 70B | 63.48 |
| 15 | Llama-3.1-Tulu-3-8B-SFT | 62.16 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-false-premise-kuq-false-assumptions-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.