A synthetic-data-driven method trains lightweight classifiers to flag user prompts as off-topic relative to a system prompt, outperforming heuristics and partially transferring to jailbreak and harmful content.
Custom categories in azure ai content safety --- learn.microsoft.com
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Flexible Large Language Models Guardrail Development Methodology Applied to Off-Topic Prompt Detection
A synthetic-data-driven method trains lightweight classifiers to flag user prompts as off-topic relative to a system prompt, outperforming heuristics and partially transferring to jailbreak and harmful content.