conftrace_

harmful instruction

2 papers

Explore in graph

Co-occurring keywords · papers mentioning both

capability degradation (1) harmful instruction defense (1) browser agent (1) model defense (1) large language model (1) agent safety (1) malicious fine-tuning (1) model capability (1) large language model alignment (1) fine-tuning attack (1)

Papers

Aligned LLMs Are Not Aligned Browser Agents ICLR 2025 SDD: Self-Degraded Defense against Malicious Fine-tuning ACL 2025