AbstentionBench - underspecified intent - SituatedQA/Geo - F1: leaderboard
Metric: Abstention F1 (%). Source: github.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 70.99 |
| 2 | O1 | 66.44 |
| 3 | Llama 3.3 70B Instruct | 64.46 |
| 4 | Llama-3.1-Tulu-3-70B-DPO | 64.04 |
| 5 | Llama-3.1-Tulu-3-70B | 63.28 |
| 6 | Llama 3.1 8B Instruct | 63.08 |
| 7 | Llama-3.1-Tulu-3-8B | 62.66 |
| 8 | Llama-3.1-Tulu-3-8B-DPO | 60.53 |
| 9 | Mistral 7B Instruct (v0.3) | 59.85 |
| 10 | Qwen 2.5 32B Instruct | 58.68 |
| 11 | Llama-3.1-Tulu-3-8B-SFT | 56.72 |
| 12 | Llama 3.1 405B Instruct | 54.23 |
| 13 | Llama 3.1 70B Instruct | 54.18 |
| 14 | DeepSeek R1 Distill Llama 70B | 53.48 |
| 15 | Gemini 1.5 Pro | 52.59 |
Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-intent-situatedqa-geo-f1 · How It Works · Data refreshed daily, snapshot 2026-09-19.