© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 71656
    Prev
    Next
    INDEX
    Explanations

    reverse direction or opposite

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

    The neuron strongly activates on code‐style punctuation and keywords (e.g. braces “{”, semicolons “;”, colons in property declarations, CSS identifiers like “background­image”), so it’s essentially detecting snippets of programming or CSS syntax rather than normal prose.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    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
     palpit
    -1.01
     nimmer
    -1.01
     beina
    -0.92
    Élet
    -0.92
    tismo
    -0.90
    quê
    -0.88
     heilig
    -0.87
     betek
    -0.87
    modelName
    -0.86
     enjeux
    -0.84
    POSITIVE LOGITS
     reverse
    1.16
     left
    1.11
     male
    1.10
     противополо
    1.08
     clockwise
    1.08
     opposite
    1.02
     Reverse
    0.99
     positive
    0.98
    left
    0.98
     reversed
    0.96
    Activations Density 0.099%

    No Known Activations