Datasets:
README: 8-way comparison section
Browse files
README.md
CHANGED
|
@@ -129,3 +129,10 @@ overlapping real tracks (all YouTube-derived); 5 further real tracks became unre
|
|
| 129 |
`v1.1/` ships the purged partition (n = 2,224), a per-track status CSV, and official v1.1
|
| 130 |
result bounds. **v1 files are unchanged** — results computed on v1 remain reproducible.
|
| 131 |
See `v1.1/RESULTS_v1.1.md`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 129 |
`v1.1/` ships the purged partition (n = 2,224), a per-track status CSV, and official v1.1
|
| 130 |
result bounds. **v1 files are unchanged** — results computed on v1 remain reproducible.
|
| 131 |
See `v1.1/RESULTS_v1.1.md`.
|
| 132 |
+
|
| 133 |
+
|
| 134 |
+
## 8-way public model comparison (2026-07-04)
|
| 135 |
+
|
| 136 |
+
Extends the v1.1 evaluation with five more publicly-available detectors on the
|
| 137 |
+
purged test partition. ArtifactNet ranks first among all eight (parameter count
|
| 138 |
+
does not predict F1 — see notes). Full table: `v1.1/RESULTS_8WAY.md`.
|