Hidden-state activations from google/gemma-3-27b-it, recorded while it lied or told the truth and was then asked whether it had lied. Part of a SPAR replication of Anthropic's post, built on its released dataset Each tensor is labelled with independently checked ground truth (did gemma actually lie on this row?), not the dataset's own didlie column, which is unusable for gemma. unanswerable, capturesecret, alibi, and the MASK belief subtypes maskknownfacts, maskstatistics, maskcontinuations, maskdisinformation (their first-run state is the pressured prompt: firstrun/mask / pressuredpredecision.pt). After gemma's answer, it is asked whether it lied, with A = yes and B = no: - Shape [63…
Organization
Model Hermeneutics
model-hermeneutics
Models in Library0
Datasets in Library1
Models on Hugging Face—
Followers5
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.