AbstentionBench - underspecified context - BB/Disambiguate - Recall Leaderboard · The Aggregate

AbstentionBench - underspecified context - BB/Disambiguate - Recall: leaderboard

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

Top models

#	Model	Score
1	Llama 3.3 70B Instruct	97.33
2	Llama 3.1 70B Instruct	96
3	Llama 3.1 70B	92
4	GPT-4o	89.33
5	Qwen 2.5 32B Instruct	89.33
6	Llama 3.1 405B Instruct	88
7	O1	84
8	Gemini 1.5 Pro	82.67
9	Llama 3.1 8B Instruct	78.67
10	Llama 3.1 8B	40
11	Llama-3.1-Tulu-3-70B-DPO	34.67
12	DeepSeek R1 Distill Llama 70B	26.67
13	Llama-3.1-Tulu-3-70B	21.33
14	Llama-3.1-Tulu-3-8B	20
15	Llama-3.1-Tulu-3-8B-DPO	16

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