© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 6737
    Prev
    Next
    INDEX
    Explanations

    Okay, here's

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

    The neuron is picking up on discourse-transition words and phrases (like “So,” “Also,” “And,” “asked,” “‘s,” “range”) that signal explanatory or sequential steps.

    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
    Wins
    0.56
     aank
    0.54
     PYTHON
    0.54
     hyperlink
    0.53
    ⌀
    0.53
     amazon
    0.50
    python
    0.50
     주요
    0.50
     coriander
    0.49
    ｷ
    0.49
    POSITIVE LOGITS
    bmatrix
    0.57
    讳
    0.54
    ressen
    0.53
     கலை
    0.52
    umen
    0.51
    кої
    0.51
    īgi
    0.51
    itôt
    0.50
    ulty
    0.49
     我們
    0.49
    Activations Density 0.015%

    No Known Activations