AbstentionBench - subjective - CoCoNot/Subjective - Recall Leaderboard · The Aggregate

AbstentionBench - subjective - CoCoNot/Subjective - Recall: leaderboard

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

Top models

#	Model	Score
1	Llama-3.1-Tulu-3-70B-DPO	94.67
2	Llama-3.1-Tulu-3-70B	92
3	Llama-3.1-Tulu-3-8B-DPO	86.67
4	Gemini 1.5 Pro	81.33
5	Llama-3.1-Tulu-3-8B	80
6	GPT-4o	78.67
7	O1	78.67
8	Llama 3.1 8B Instruct	76
9	Llama 3.1 405B Instruct	73.33
10	Qwen 2.5 32B Instruct	70.67
11	Llama-3.1-Tulu-3-8B-SFT	70.67
12	Llama 3.1 70B Instruct	69.33
13	Mistral 7B Instruct (v0.3)	65.33
14	Llama 3.1 70B	61.33
15	Llama 3.3 70B Instruct	58.67

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