AbstentionBench - underspecified context - Musique - Recall: leaderboard

Metric: Abstention Recall (%). Source: github.com. 20 models tracked.

Top models

#ModelScore
1Qwen 2.5 32B Instruct87.08
2Mistral 7B Instruct (v0.3)86.4
3Gemini 1.5 Pro83.51
4Llama 3.1 8B Instruct80.85
5Llama 3.3 70B Instruct78.34
6Llama 3.1 405B Instruct76.98
7Llama 3.1 70B Instruct73.4
8GPT-4o62.77
9Llama 3.1 8B54.71
10Llama-3.1-Tulu-3-70B-DPO54.18
11Llama 3.1 70B51.52
12O148.56
13Llama-3.1-Tulu-3-70B47.19
14DeepSeek R1 Distill Llama 70B36.25
15Llama-3.1-Tulu-3-8B-DPO18.54

Interactive version: theaggregate.ai/benchmark?slug=abstentionbench-underspecified-context-musique-recall · How It Works · Data refreshed daily, snapshot 2026-09-19.