AbstentionBench - false premise - CoCoNot/False presumptions - Precision: leaderboard
Metric: Abstention Precision (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek R1 Distill Llama 70B | 98.21 |
| 2 | Llama-3.1-Tulu-3-8B-SFT | 98.15 |
| 3 | Llama-3.1-Tulu-3-70B | 97.1 |
| 4 | GPT-4o | 96.88 |
| 5 | Llama-3.1-Tulu-3-70B-DPO | 95.65 |
| 6 | Llama 3.3 70B Instruct | 95.31 |
| 7 | Qwen 2.5 32B Instruct | 93.94 |
| 8 | Llama-3.1-Tulu-3-8B | 93.75 |
| 9 | Mistral 7B Instruct (v0.3) | 93.65 |
| 10 | Llama 3.1 70B Instruct | 93.55 |
| 11 | Llama-3.1-Tulu-3-8B-DPO | 91.3 |
| 12 | O1 | 91.25 |
| 13 | Gemini 1.5 Pro | 90.63 |
| 14 | Llama 3.1 405B Instruct | 88 |
| 15 | Llama 3.1 8B Instruct | 86.36 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-false-premise-coconot-false-presumptions-precision · How It Works · Data refreshed daily, snapshot 2026-09-19.