Multi-label Text Classification on WM Nicol
94.67Micro-F1ProtoNet + RAPT#
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ProtoNet + RAPT#Category=best ML baseline2026.05 | 94.67 | 87.04 | — | — | — | — | |
| DeBERTa-v3-large + RAPT#Category=best transformer baseline, Evaluation Protocol=fully fine-tuned2026.05 | 85.12 | 79.57 | — | — | — | — | |
| Qwen3-32BPrompting Strategy=Zero-shot (Semantic), Semantic Labels Provided=true2026.05 | 61.34 | 52.2 | -35.2 | -40 | -27.9 | -34.4 | |
| Qwen3-32BPrompting Strategy=Few-shot (Anon.), Few-shot K=52026.05 | 44.08 | 37.45 | -53.4 | -57 | -48.2 | -52.9 | |
| Qwen3-32BPrompting Strategy=Few-shot (Anon.), Few-shot K=32026.05 | 42.32 | 36.48 | -55.3 | -58.1 | -50.3 | -54.2 | |
| Qwen3-32BPrompting Strategy=Few-shot (Anon.), Few-shot K=12026.05 | 9.96 | 10.57 | -89.5 | -87.9 | -88.3 | -86.7 | |
| Qwen3-32BPrompting Strategy=Zero-shot (Anonymized)2026.05 | 9.58 | 8.45 | -89.9 | -90.3 | -88.7 | -89.4 |