A single linear direction in an observer model's residual stream detects contextual hallucinations, transfers across models and datasets, and causally steers generation hallucination rates.
J., Gurnee, W., and Tegmark, M
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Single Direction of Truth: An Observer Model's Linear Residual Probe Exposes and Steers Contextual Hallucinations
A single linear direction in an observer model's residual stream detects contextual hallucinations, transfers across models and datasets, and causally steers generation hallucination rates.