© 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
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 129891
    Prev
    Next
    INDEX
    Explanations

    The neuron fires on emotionally charged content words—especially verbs and nouns that convey strong judgments, reactions, or vivid actions (e.g. “being salty,” “hope they lose,” “cheering when people die,” “grow,” “shri­veling,” “resentment”).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    succeeding or failing

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/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
    잊
    -0.93
    mettere
    -0.88
     truer
    -0.84
     alguns
    -0.83
     allerede
    -0.82
    뻤
    -0.82
    '||
    -0.81
    ceria
    -0.81
     mediados
    -0.81
     siap
    -0.79
    POSITIVE LOGITS
     success
    1.51
     succeed
    1.41
     successful
    1.37
    uccess
    1.28
    success
    1.13
     Success
    1.11
     succeeds
    1.09
     SUCCESS
    1.08
     successes
    1.07
    succeed
    1.03
    Activations Density 0.037%

    No Known Activations