Optimizing input embeddings sub-lexically via black-box zeroth-order gradients neutralizes all safety-flagged responses from aligned models on standard benchmarks.
Training language models to follow instructions with human 14 feedback
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Test-Time Safety Alignment
Optimizing input embeddings sub-lexically via black-box zeroth-order gradients neutralizes all safety-flagged responses from aligned models on standard benchmarks.