ArgBench (Chain-of-Thought): 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 chain-of-thought 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, chain-of-thought prompting; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-4.157.7#505
2Llama 3.3 70B Instruct52.9#1385
3Qwen 3 32B52.2#972
4Mixtral 8x7B Instruct50.1#4696
5Mistral-Small-Instruct-240949.7#2016
6Llama 3.1 8B Instruct49.4#4600
7Mistral 7B Instruct (v0.3)48.3#6750
8DeepSeek R1 Distill Qwen 32B47.7#1834
9Phi-mini-MoE-instruct47.2
10Qwen 3 4B47.1#2348
11DeepSeek-R1-Distill-Qwen-7B45.4#8055

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=argbench-chain-of-thought · How It Works · Data refreshed daily, snapshot 2026-11-21.