UnrestSent200K - Few-Shot: leaderboard
Sentiment of Bangla social-media comments during the 2024 Bangladesh uprising, few-shot with expert Bangla-native prompting. Only prompted LLMs are published. F1, higher is better. Source: arxiv.org/abs/2609.16997.
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini | 75 | #598 |
| 2 | GPT-4.1 Mini | 71 | #363 |
| 3 | GPT-4.1 Nano | 61 | #730 |
| 4 | Qwen 2.5 7B | 60 | #895 |
| 5 | Qwen 3 4B Instruct | 57 | #661 |
| 6 | Qwen 3 4B | 56 | #834 |
| 7 | Llama 3.1 8B | 55 | #1130 |
| 8 | Gemma 3 4B | 54 | #1088 |
| 9 | Qwen 2.5 3B | 52 | #1177 |
| 10 | Gemma 3 1B | 47 | #1505 |
| 11 | Qwen 2.5 1.5B | 40 | #1412 |
| 12 | Llama 3.2 3B | 33 | #1376 |
| 13 | Qwen 2.5 0.5B | 23 | #1536 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=unrestsent200k-few-shot · How It Works · Data refreshed daily, snapshot 2026-11-05.