Final-token probes miss distributed unsafe evidence in jailbreaks, but a PCA-HMM model on prefill trajectories recovers many misses without naive pooling's false positives.
org/abs/2511.14195
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Before the Last Token: Diagnosing Final-Token Safety Probe Failures
Final-token probes miss distributed unsafe evidence in jailbreaks, but a PCA-HMM model on prefill trajectories recovers many misses without naive pooling's false positives.