© 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-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 9443
    Prev
    Next
    INDEX
    Explanations

    can, could, will

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

    The neuron is tuned to pick up on modal/causative verbs (can, could, would, cause, lead, etc.) that signal potential risks or negative outcomes (warnings).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    drawSprites
    -1.49
     crucial
    -1.45
     apparaître
    -1.41
    cking
    -1.37
    
    -1.36
     boire
    -1.28
     этого
    -1.28
     baden
    -1.27
    helping
    -1.26
    '&:
    -1.25
    POSITIVE LOGITS
     it
    1.54
     while
    1.52
     you
    1.52
     or
    1.49
     terlalu
    1.48
     mientras
    1.45
     слишком
    1.40
     there
    1.38
    直
    1.35
     too
    1.34
    Activations Density 0.090%

    No Known Activations