© 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-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 73104
    Prev
    Next
    INDEX
    Explanations

    code blocks and mathematical expressions

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

    The neuron strongly activates on programming‐style tokens—especially function or method calls (identifiers followed by “()”) and code directives (like “v-”)—i.e. it flags snippets of code syntax.

    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
    either
    -0.91
     แท
    -0.84
     podrían
    -0.84
    instead
    -0.83
    ′
    -0.83
    昀
    -0.83
    (;;)
    -0.82
     Какой
    -0.82
    这个时候
    -0.82
     starvation
    -0.82
    POSITIVE LOGITS
    ншот
    0.82
     subsequently
    0.82
    Ẫ
    0.77
     completed
    0.76
     району
    0.76
    mitives
    0.75
     bär
    0.75
     steadily
    0.74
     every
    0.73
     repeatedly
    0.73
    Activations Density 0.020%

    No Known Activations