AbstentionBench - underspecified context - QASPER - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama-3.1-Tulu-3-8B-DPO | 44.51 |
| 2 | Llama-3.1-Tulu-3-8B | 44.31 |
| 3 | Llama-3.1-Tulu-3-70B-DPO | 41.85 |
| 4 | Qwen 2.5 32B Instruct | 41.4 |
| 5 | DeepSeek R1 Distill Llama 70B | 41.03 |
| 6 | Llama 3.3 70B Instruct | 40.97 |
| 7 | Llama 3.1 8B Instruct | 40.45 |
| 8 | Llama-3.1-Tulu-3-70B | 40 |
| 9 | Llama 3.1 405B Instruct | 39.69 |
| 10 | Mistral 7B Instruct (v0.3) | 39.57 |
| 11 | GPT-4o | 38 |
| 12 | Llama 3.1 70B Instruct | 37.37 |
| 13 | Llama-3.1-Tulu-3-8B-SFT | 35.33 |
| 14 | Gemini 1.5 Pro | 35.29 |
| 15 | Llama 3.1 70B | 33.74 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-qasper-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.