Cultural norm reasoning on NormAd South Asia 1.0 (test)
77.8AccuracyQwen3.5-4B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3.5-4BPrompt Language=Source-Language Stories2026.03 | 77.8 | |
| Tiny Aya EarthPrompt Language=Source-Language Stories2026.03 | 70.47 | |
| Tiny Aya WaterPrompt Language=Source-Language Stories2026.03 | 69.6 | |
| Tiny Aya GlobalPrompt Language=Source-Language Stories2026.03 | 69.07 | |
| Gemma3-4BPrompt Language=Source-Language Stories2026.03 | 67.77 | |
| Tiny Aya FirePrompt Language=Source-Language Stories2026.03 | 67.77 | |
| Qwen3-4BPrompt Language=Source-Language Stories2026.03 | 65.97 | |
| SmolLM3-3BPrompt Language=Source-Language Stories2026.03 | 64.97 | |
| Ministral-3-3BPrompt Language=Source-Language Stories2026.03 | 57.03 |