Jet tagging on HLS4ML 16-feature (test)
95.99AUC (Gluon)JetFormer
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| JetFormerConstituents=100, Params=162,843, FLOPs=14M2026.01 | 95.99 | 83.05 | 93.92 | 98.12 | 97.83 | 97.44 | |
| JetFormerConstituents=150, Params=238,443, FLOPs=24M2026.01 | 95.96 | 82.97 | 93.89 | 98.1 | 97.81 | 97.42 | |
| JetFormerConstituents=50, Params=117,243, FLOPs=6M2026.01 | 95.9 | 82.88 | 93.88 | 98.09 | 97.8 | 97.43 | |
| JetFormerConstituents=30, Params=107,403, FLOPs=4M2026.01 | 95.34 | 81.77 | 93.44 | 97.68 | 97.36 | 97.15 | |
| JEDI-netConstituents=100, Params=33,625, FLOPs=116M2026.01 | 95.29 | — | 93.01 | 97.39 | 96.79 | 96.83 | |
| JEDI-net w/ Σ OConstituents=150, Params=8,767, FLOPs=458M2026.01 | 95.28 | — | 92.9 | 96.95 | 96.49 | 96.77 |