AbstentionBench - underspecified intent - KUQ/Ambiguous - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 76.72 |
| 2 | Gemini 1.5 Pro | 74.88 |
| 3 | O1 | 72.07 |
| 4 | Llama-3.1-Tulu-3-70B-DPO | 70.12 |
| 5 | Llama-3.1-Tulu-3-70B | 69.7 |
| 6 | Qwen 2.5 32B Instruct | 69.52 |
| 7 | Llama 3.3 70B Instruct | 69.47 |
| 8 | Llama-3.1-Tulu-3-8B-DPO | 68.74 |
| 9 | Llama-3.1-Tulu-3-8B | 68.67 |
| 10 | Mistral 7B Instruct (v0.3) | 66.55 |
| 11 | Llama 3.1 70B Instruct | 65.3 |
| 12 | Llama-3.1-Tulu-3-8B-SFT | 64.71 |
| 13 | Llama 3.1 405B Instruct | 64.18 |
| 14 | Llama 3.1 8B Instruct | 61.3 |
| 15 | Llama 3.1 70B | 60.21 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-intent-kuq-ambiguous-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.