ArgBench (Few-Shot) - Argument Mining: leaderboard

ArgBench (2026) unifies 46 computational-argumentation tasks from 33 existing datasets into five skills. This board is the macro F1 averaged over the argument mining tasks under few-shot prompting, times 100; higher is better.

Metric: Macro F1 (0-1, times 100) averaged over the argument mining (unit segmentation and classification) tasks, few-shot prompting; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-4.159.7#506
2Llama 3.3 70B Instruct52.7#1394
3Qwen 3 32B50.1#983
4Mixtral 8x7B Instruct44.3#4692
5Mistral 7B Instruct (v0.3)43.4#6759
6Qwen 3 4B40.6#2379
7Mistral-Small-Instruct-240940#2026
8Llama 3.1 8B Instruct36.7#4655
9DeepSeek R1 Distill Qwen 32B33.3#1844
10Phi-mini-MoE-instruct32.5
11DeepSeek-R1-Distill-Qwen-7B30.8#7976

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=argbench-few-shot-argument-mining · How It Works · Data refreshed daily, snapshot 2026-11-22.