Intent Classification on MASSIVE W5H2
0Cost/1KRule-based (W5H2)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Rule-based (W5H2)Training=0 examples2026.02 | 0 | 60.7 | — | — | |
| SetFit-Multi (118M)Training=64 examples2026.02 | 0 | 84.4 | 5 | — | |
| SetFit-EN (22M)Training=64 examples2026.02 | 0 | — | 25 | 0.813 | |
| BERT-base (supervised)Training=full train2026.02 | 0 | 97.3 | 85 | 0.926 | |
| LLM 0-shot (GPT-oss-20b)Training=0 examples2026.02 | 0.089 | 68.8 | — | 0.65 | |
| LLM 2-shotTraining=16 examples2026.02 | 0.11 | 75 | — | — | |
| LLM 4-shotTraining=32 examples2026.02 | 0.13 | 79 | — | — |