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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 72007
    Prev
    Next
    INDEX
    Explanations

    names and ages

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

    The neuron fires on “weird” or out‐of‐place tokens – things like table‐markup bits (| colspan=…), standalone numbers, or other rare/garbage subwords – i.e. low-frequency, oddball tokens.

    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
     its
    -1.28
     about
    -1.23
     several
    -1.18
     almost
    -1.12
     nearly
    -1.07
     několika
    -1.02
     it
    -1.01
     lutar
    -0.99
     dozens
    -0.98
     actually
    -0.97
    POSITIVE LOGITS
    ktive
    1.22
    誦
    1.21
     خانه
    1.20
    Comté
    1.20
     учащихся
    1.14
    Barcelona
    1.05
    ことを
    1.05
    siger
    1.05
    ب
    1.05
    lées
    1.04
    Activations Density 0.001%

    No Known Activations