© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 1246
    Prev
    Next
    INDEX
    Explanations

    acknowledging when a model is not GPT

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

    The neuron strongly activates on discourse‐structuring connectors and transition cues (e.g. “Since,” “While,” “if,” “which,” “for example”), signaling that it’s looking for words or phrases that mark logical or rhetorical links 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
     근데
    0.71
     которая
    0.61
     которые
    0.59
    但是我
    0.58
    M
    0.56
     但是
    0.55
     которую
    0.54
    但是
    0.54
     लेकिन
    0.53
     когда
    0.53
    POSITIVE LOGITS
     policymakers
    0.57
     anthropologists
    0.54
     acknowledging
    0.54
     이러한
    0.53
     individui
    0.53
    これらの
    0.52
     lawmakers
    0.51
     archaeologists
    0.51
     precludes
    0.50
     novelists
    0.50
    Activations Density 0.616%

    No Known Activations