A two-part 'audit then review then code' prompt stops code LLMs from copying harmful words into identifiers, while a new 12-level score measures the correctness-safety trade-off.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Think Before You Code: Dual Reasoning for the NLSafety-Utility Trade-Off in LLM Code Generation
A two-part 'audit then review then code' prompt stops code LLMs from copying harmful words into identifiers, while a new 12-level score measures the correctness-safety trade-off.