A safeguard that uses speculative inference on small language models to produce draft responses for safety prediction, lowering false negatives in pre-model jailbreak detection.
Revisiting the robust alignment of circuit breakers.CoRR, abs/2407.15902
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3representative citing papers
JBShield is vulnerable to adaptive JB-GCG attacks (up to 53% ASR) because jailbreak representations occupy a distinct region in refusal-direction space; the new RTV defense using Mahalanobis detection on multi-layer fingerprints reaches 0.99 AUROC and limits adaptive ASR to 7%.
Harmful LLM outputs rely on a compact, cross-harm set of weights distinct from benign skills; alignment compresses them, and pruning them reduces emergent misalignment.
citing papers explorer
-
Exploring and Developing a Pre-Model Safeguard with Draft Models
A safeguard that uses speculative inference on small language models to produce draft responses for safety prediction, lowering false negatives in pre-model jailbreak detection.
-
Revisiting JBShield: Breaking and Rebuilding Representation-Level Jailbreak Defenses
JBShield is vulnerable to adaptive JB-GCG attacks (up to 53% ASR) because jailbreak representations occupy a distinct region in refusal-direction space; the new RTV defense using Mahalanobis detection on multi-layer fingerprints reaches 0.99 AUROC and limits adaptive ASR to 7%.
-
Large Language Models Generate Harmful Responses Using a Distinct Mechanism, Shared Across Harm Types
Harmful LLM outputs rely on a compact, cross-harm set of weights distinct from benign skills; alignment compresses them, and pruning them reduces emergent misalignment.