© 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
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 4683
    Prev
    Next
    INDEX
    Explanations

    unintended consequences or errors

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

    The neuron is essentially responding to mid‐frequency, content‐bearing words (i.e. relatively “rare” nouns and verbs) rather than common function words.

    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
    💙
    1.04
    💚
    0.96
    💫
    0.94
    💪
    0.91
     vivió
    0.91
    🦋
    0.90
    🌺
    0.89
    天使
    0.88
    ☀️
    0.88
    ✨
    0.88
    POSITIVE LOGITS
     misleading
    1.44
     unduly
    1.35
     inaccurate
    1.31
     unrealistic
    1.31
     erroneous
    1.27
     insufficient
    1.25
     inadvert
    1.23
     unnecessarily
    1.20
     unreasonable
    1.19
     inconvenient
    1.18
    Activations Density 0.871%

    No Known Activations