© 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. 31-GEMMASCOPE-2-RES-16K
    4. 475
    Prev
    Next
    INDEX
    Explanations

    The neuron fires strongly on self‐referential language, especially first‐person pronouns and adjoining verbs (e.g. “I,” “can,” “tell,” “say,” “feel”) signaling a personal statement.

    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_31_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
     durante
    0.35
    .,
    0.32
     ovvero
    0.31
     №
    0.30
     необхідно
    0.30
     considerando
    0.30
     which
    0.29
    .
    0.29
     cercando
    0.29
     during
    0.29
    POSITIVE LOGITS
     easily
    0.61
    easily
    0.56
     afford
    0.55
     be
    0.51
     facilement
    0.51
    opies
    0.49
    轻易
    0.48
     withstand
    0.48
     conceivably
    0.47
     इजीली
    0.46
    Activations Density 1.011%

    No Known Activations