Cross-Lingual Tamper Resistance (Datasets) Collection Per-language safety-eval datasets for the cross-lingual tamper-resistance study on Tiny Aya Fire. • 4 items • Updated about 4 hours ago
Cross-Lingual Tamper Resistance (Models) Collection Bengali-attacked Tiny Aya Fire, repaired post-hoc by low-rank projection of the attack's weight update (Exp A, k=1). • 1 item • Updated about 4 hours ago
Cross-Lingual Tamper Resistance (Datasets) Collection Per-language safety-eval datasets for the cross-lingual tamper-resistance study on Tiny Aya Fire. • 4 items • Updated about 4 hours ago