© 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-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 15180
    Prev
    Next
    INDEX
    Explanations

    parts and phases

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

    The neuron isn’t looking for any specific word pattern at all but simply picks out the single highest‐activation token in each input chunk—i.e. it’s acting like a max‐pool over its inputs.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_24_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    a
    0.80
    sizeof
    0.76
    Again
    0.74
    కే
    0.74
    or
    0.73
    NaOH
    0.72
    ケート
    0.72
    స
    0.72
    ransport
    0.71
    ヘ
    0.70
    POSITIVE LOGITS
     grime
    0.83
     şeyler
    0.80
     shard
    0.77
     estuary
    0.76
     grinder
    0.75
     nito
    0.75
    ાઓ
    0.73
     savvy
    0.73
     wondrous
    0.73
     legible
    0.73
    Activations Density 0.000%

    No Known Activations