activity
20242026
most citedReleasing Malevolence from Benevolence: The Menace of Benign Data on Machine Unlearning

2 citations · 2 across the 2 of their papers we have counts for

collaborators
Showing cs.LGShow all

1 paper · 1 filter

cs.LG2026

ProbGuard: Calibrated Safety Risk Estimation from LLM Output Distributions

Xinzhe Huang, Biwu Yao, Kedong Xiu +4

Recent research on Large Language Model (LLM) safety has widely adopted guardrails to identify unsafe LLM outputs. Existing guardrails typically formulate safety assessment as a de…