cvc-mmu/clip-vit-b-16-salmu-compromised

Zero-Shot Image Classification • Updated Mar 30 • 4 • 1

Note Compromised CLIP ViT-B/16 model used as the starting point for unlearning.

cvc-mmu/clip-vit-b-16-salmu-clean

Zero-Shot Image Classification • Updated Mar 30 • 3

Note Clean reference CLIP ViT-B/16 model trained without sensitive associations.

Note Main SALMUBench evaluation dataset with forget, holdout, and retain splits.

Note Training dataset used to create the Compromised model in SALMUBench experiments.

Note Benchmark splits derived from the original dataset before SALMUBench redistribution.

Note Original dataset before redistribution, provided for custom splits or retraining.