AbstentionBench - underspecified context - BB/Disambiguate - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.1 70B | 66.99 |
| 2 | Llama 3.1 405B Instruct | 62.56 |
| 3 | Llama 3.1 70B Instruct | 61.28 |
| 4 | Qwen 2.5 32B Instruct | 61.19 |
| 5 | Llama 3.3 70B Instruct | 60.58 |
| 6 | O1 | 60 |
| 7 | Gemini 1.5 Pro | 53.68 |
| 8 | GPT-4o | 50.76 |
| 9 | Llama 3.1 8B Instruct | 46.83 |
| 10 | Llama 3.1 8B | 45.8 |
| 11 | Llama-3.1-Tulu-3-70B-DPO | 44.83 |
| 12 | DeepSeek R1 Distill Llama 70B | 35.71 |
| 13 | Llama-3.1-Tulu-3-70B | 32.99 |
| 14 | Llama-3.1-Tulu-3-8B | 29.13 |
| 15 | Llama-3.1-Tulu-3-8B-DPO | 23.76 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-bb-disambiguate-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.