© 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. 13268
    Prev
    Next
    INDEX
    Explanations

    be brutally honest

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

    The neuron detects first‐person opinion or commentary phrases (e.g. “I…,” “to be honest,” “I want to…”).

    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
    tyard
    0.57
    ilver
    0.54
     тысячи
    0.53
     temporal
    0.53
    waitFor
    0.50
     verify
    0.49
     Bloomington
    0.49
    })^{
    0.49
    indungan
    0.49
     rote
    0.48
    POSITIVE LOGITS
     honestly
    2.45
     honest
    2.28
     truth
    2.22
     honesty
    2.06
     Honestly
    2.06
    truth
    2.05
     truthfully
    2.00
    honest
    1.99
     tbh
    1.98
     frankly
    1.96
    Activations Density 0.179%

    No Known Activations