© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 40087
    Prev
    Next
    INDEX
    Explanations

    natural, abnormal, or specific features

    np_acts-logits-general · gemini-2.5-flash-lite

    The neuron flags unusually informative, domain‐specific content words—especially technical or diagnostic terms (e.g. “hypoglycaemia,” “vascularity,” “angiographically”) as well as question‐forming cues (“ask,” “?,” “for instance”).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    thenia
    -1.02
     MÉXICO
    -0.98
    Manfaat
    -0.97
     already
    -0.92
    ؋
    -0.90
    ּוֹ
    -0.89
     manfaat
    -0.88
    تهى
    -0.88
    няя
    -0.87
    ύ
    -0.86
    POSITIVE LOGITS
    Their
    1.19
    彼らの
    1.09
     their
    1.08
     suele
    1.08
    Typically
    1.07
     обычно
    1.05
     indicative
    1.02
    Unlike
    1.02
     tenían
    1.01
    特征
    0.99
    Activations Density 0.029%

    No Known Activations