ArgBench (Few-Shot): leaderboard
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.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 63.3 |
| 2 | Llama 3.3 70B Instruct | 58.7 |
| 3 | Qwen 3 32B | 57 |
| 4 | Mixtral 8x7B Instruct (v0.1) | 53.5 |
| 5 | Qwen 3 4B | 49.9 |
| 6 | Mistral-Small-Instruct-2409 | 49.7 |
| 7 | Llama 3.1 8B Instruct | 47.1 |
| 8 | Mistral 7B Instruct (v0.3) | 45.8 |
| 9 | DeepSeek R1 Distill Qwen 32B | 43.8 |
| 10 | DeepSeek-R1-Distill-Qwen-7B | 41.5 |
Interactive version: theaggregate.ai/benchmark?slug=argbench-few-shot · How It Works · Data refreshed daily, snapshot 2026-10-07.