Red-team bench: 30 indirect prompt-injection scenarios (22 attacks, 8 benign), SafeTune set, Tiny Aya transcripts. ASR 54.5→36.4%, over-refusal 12.5%
Muhamad Viggo Pratama Luthfiansyah
binasalama12
AI & ML interests
None yet
Recent Activity
updated a collection 9 days ago
Trust No Tool Output — Tiny Aya Agent Red-Team Bench (Lexsi) updated a collection 9 days ago
Trust No Tool Output — Tiny Aya Agent Red-Team Bench (Lexsi) updated a model 9 days ago
binasalama12/tiny-aya-steering-lexsiOrganizations
None yet