STREAM fine-tunes a small reasoning model on human-labeled, reason-annotated multi-turn dialogues and uses it to warn target LLMs, cutting average attack success rates by roughly half while keeping benchmark scores close.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SafeTy Reasoning Elicitation Alignment for Multi-Turn Dialogues
STREAM fine-tunes a small reasoning model on human-labeled, reason-annotated multi-turn dialogues and uses it to warn target LLMs, cutting average attack success rates by roughly half while keeping benchmark scores close.