© 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-4-31B
    5. 30-RES-MATRYOSHKA-131K
    7. 117255
    Prev
    Next
    INDEX
    Explanations

    **Detecting Patterns in Data**I'm currently analyzing the provided lists to pinpoint the specific patterns each neuron is designed to detect. My focus is on distilling this information into a concise, non-sentence phrase that accurately reflects the neuron's predictive behavior, as requested.Please provide the lists so I can start the analysis.**Detecting Patterns in Data**I'm ready to analyze the provided lists and explain the neuron's behavior. Please paste the four lists (<MAX_ACTIVATING_TOKENS>, <TOKENS_AFTER_MAX_ACTIVATING_TOKEN>, <TOP_POSITIVE_LOGITS>, <TOP_ACTIVATING_TEXTS>) so I can proceed

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    decoderesearch/gemma-4-saes/gemma-4-31b
    Prompts (Dashboard)
    16,384 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     arresting
    -0.06
     क्रमवारीत
    -0.06
    最终
    -0.06
    最終
    -0.06
     beside
    -0.06
     जखमी
    -0.06
     kär
    -0.06
    Arte
    -0.06
     simult
    -0.06
     distintas
    -0.05
    POSITIVE LOGITS
     memiliki
    0.07
    ọi
    0.06
     vying
    0.06
    慫
    0.06
     прошлом
    0.06
     PRP
    0.06
     активно
    0.06
     Bereich
    0.06
     praz
    0.06
    专属
    0.06
    Activations Density 0.003%

    No Known Activations