© 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. 75
    Prev
    Next
    INDEX
    Explanations

    The neuron fires almost exclusively on the “<bos>” token (and other markers of new segments), so it’s detecting the start-of-sequence or sentence/document boundary.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    prepositions indicating relationships, particularly "of" and "to", in formal or technical writing.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    prepositions and related function words that mark relationships between ideas or actions (e.g., “of,” “on,” “to,” “in,” “for,” “between”).

    oai_token-act-pair · gpt-5Triggered 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
    Ammo
    0.46
    Cache
    0.38
    TAX
    0.36
    also
    0.36
    Rail
    0.36
    RootDir
    0.36
    I
    0.35
    Trailer
    0.34
    ﻤ
    0.34
     ตุลาคม
    0.34
    POSITIVE LOGITS
     these
    0.93
     aceste
    0.83
     this
    0.80
     such
    0.79
    这类
    0.78
     acest
    0.78
     이러한
    0.73
    こうした
    0.70
     diesen
    0.65
     thefe
    0.65
    Activations Density 14.067%

    No Known Activations