ArgBench (Few-Shot): leaderboard

ArgBench (2026) unifies 46 computational-argumentation tasks from 33 existing datasets into five skills. This board is the macro average over the five skills (mining, perspective assessment, quality assessment, reasoning, and generation scored by BERTScore) under few-shot prompting, from 0 to 1 times 100; higher is better.

Metric: Macro average (0-1 F1-type scores, times 100) over the five skills (mining, perspective, quality, reasoning, and generation scored by BERTScore) of 46 tasks from 33 existing argumentation datasets, 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-4.163.3#506
2Llama 3.3 70B Instruct58.7#1394
3Qwen 3 32B57#983
4Mixtral 8x7B Instruct53.5#4692
5Qwen 3 4B49.9#2379
6Mistral-Small-Instruct-240949.7#2026
7Llama 3.1 8B Instruct47.1#4655
8Mistral 7B Instruct (v0.3)45.8#6759
9Phi-mini-MoE-instruct45.1
10DeepSeek R1 Distill Qwen 32B43.8#1844
11DeepSeek-R1-Distill-Qwen-7B41.5#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 · How It Works · Data refreshed daily, snapshot 2026-11-22.