AbstentionBench - underspecified context - WorldSense - Precision: leaderboard
Metric: Abstention Precision (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O1 | 89.25 |
| 2 | Llama 3.1 405B Instruct | 87.48 |
| 3 | DeepSeek R1 Distill Llama 70B | 86.18 |
| 4 | Gemini 1.5 Pro | 85.81 |
| 5 | GPT-4o | 84.83 |
| 6 | Qwen 2.5 32B Instruct | 77.27 |
| 7 | Llama 3.1 70B | 76.86 |
| 8 | Llama 3.1 8B | 67.76 |
| 9 | Llama 3.1 70B Instruct | 66.7 |
| 10 | Llama 3.3 70B Instruct | 65.2 |
| 11 | Llama-3.1-Tulu-3-70B | 51.75 |
| 12 | Llama-3.1-Tulu-3-70B-DPO | 51.17 |
| 13 | Mistral 7B Instruct (v0.3) | 40.91 |
| 14 | Llama 3.1 8B Instruct | 40.34 |
| 15 | Llama-3.1-Tulu-3-8B | 39.69 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-worldsense-precision · How It Works · Data refreshed daily, snapshot 2026-09-19.