Open-vocabulary sensor-language retrieval on ActivityNarrated XS 1.0 (Cross-Subject)
38.5R@1Spectral + Q-Former + T.a.
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Spectral + Q-Former + T.a.Discretization=Spectral, Aggregation=Q-Former, T.a.=true2026.04 | 38.5 | 71.1 | 59 | 68 | |
| Spectral + Q-FormerDiscretization=Spectral, Aggregation=Q-Former2026.04 | 34.7 | 66.2 | 53 | 63 | |
| Spectral + attentionDiscretization=Spectral, Aggregation=attention2026.04 | 31.4 | 58.9 | 44 | 58 | |
| Spectral + mean poolingDiscretization=Spectral, Aggregation=mean pooling2026.04 | 29.1 | 55.3 | 43 | 56 | |
| Raw + Q-FormerDiscretization=Raw, Aggregation=Q-Former2026.04 | 26.1 | 55.1 | 41 | 54 | |
| Raw + attentionDiscretization=Raw, Aggregation=attention2026.04 | 23.3 | 50.5 | 43 | 52 | |
| PaLM-E2026.04 | 23.1 | 50.4 | 38 | 50 | |
| Raw + mean poolingDiscretization=Raw, Aggregation=mean pooling2026.04 | 22.8 | 49.3 | 36 | 50 | |
| OVHAR2026.04 | 21.9 | 47.6 | 35 | 49 | |
| IMU2CLIP2026.04 | 10.4 | 32.7 | 21 | 33 |