Personalized Classification on LaMP-2 (val)
53.1AccuracyPHFCompass
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PHFCompassParadigm=PHF Implements2026.06 | 53.1 | 46.1 | |
| L-FParadigm=PHF Implements2026.06 | 52.8 | 49 | |
| T-H+T-FParadigm=PHF Implements2026.06 | 51.8 | 47.8 | |
| T-H+L-FParadigm=PHF Implements2026.06 | 50.6 | 47.4 | |
| PPlugParadigm=Flat: Aggregation-Based2026.06 | 50.3 | 44.8 | |
| DenseParadigm=Flat: Retrieval-Based2026.06 | 49 | 46.8 | |
| ROPGParadigm=Flat: Retrieval-Based2026.06 | 48.9 | 45.5 | |
| T-FParadigm=PHF Implements2026.06 | 46.7 | 43.3 | |
| T-HParadigm=PHF Implements2026.06 | 45.7 | 44.4 | |
| BM25Paradigm=Flat: Retrieval-Based2026.06 | 45.3 | 43.7 | |
| RecencyParadigm=Flat: Retrieval-Based2026.06 | 44.9 | 43.2 | |
| OPPUParadigm=Flat: Aggregation-Based2026.06 | 44.4 | 40 | |
| DiffMeanParadigm=Flat: Aggregation-Based2026.06 | 37.4 | 34.3 | |
| Zero-ShotBase Model=Qwen2.5-7B-Instruct2026.06 | 36.6 | 33.7 |