ArgBench (Few-Shot) - Quality Assessment: 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 quality assessment tasks under few-shot prompting, times 100; higher is better.
Metric: Macro F1 (0-1, times 100) averaged over the argument quality assessment tasks, few-shot prompting; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 47.7 | #505 |
| 2 | Qwen 3 32B | 42.2 | #972 |
| 3 | Llama 3.3 70B Instruct | 41.8 | #1385 |
| 4 | DeepSeek R1 Distill Qwen 32B | 40.2 | #1834 |
| 5 | DeepSeek-R1-Distill-Qwen-7B | 39.5 | #8055 |
| 6 | Mixtral 8x7B Instruct | 39 | #4696 |
| 7 | Qwen 3 4B | 36.7 | #2348 |
| 8 | Llama 3.1 8B Instruct | 35.8 | #4600 |
| 9 | Mistral 7B Instruct (v0.3) | 34.4 | #6750 |
| 10 | Phi-mini-MoE-instruct | 32.9 | |
| 11 | Mistral-Small-Instruct-2409 | 31.1 | #2016 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=argbench-few-shot-quality-assessment · How It Works · Data refreshed daily, snapshot 2026-11-21.