© 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
    3. Gemma-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 10441
    Prev
    Next
    INDEX
    Explanations

    produces and gives

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

    The neuron spots label‐style or heading‐style terms—i.e. category names, section titles, and role/descriptive labels in the text.

    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
    都可以
    1.12
    都会
    1.09
     deberán
    1.05
    都要
    1.02
     eivät
    1.02
    都不会
    1.01
    都需要
    0.99
    都會
    0.98
    都能
    0.97
     sollen
    0.97
    POSITIVE LOGITS
     possesses
    2.39
     operates
    2.38
     enjoys
    2.38
     employs
    2.36
     owns
    2.33
     maintains
    2.33
     provides
    2.28
     delivers
    2.28
     gives
    2.27
     knows
    2.23
    Activations Density 0.801%

    No Known Activations