© 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
    2. Qwen3-1.7B
    3. 26-LLAMASCOPE-2-LORSA-16K-K64
    4. 2278
    Prev
    Next
    INDEX
    Explanations

    <THINKING> Method 2 succeeds: All TOKENS_AFTER_MAX_ACTIVATING_TOKEN are variations of the word “digital” (e.g., digital, Digital, 数字). Method 3 confirms the same pattern with TOP_POSITIVE_LOGITS focusing on “digital”, “digit”, etc. Thus the neuron detects the concept “digital”. Method used: 2, because the consistent token after activation is the pattern. </THINKING> digital

    unknown · unknown
    New Auto-Interp
    Top Features by Cosine Similarity
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     protein
    -17.63
    熔
    -17.00
    蛋白质
    -17.00
    rotein
    -16.63
    蛋白
    -16.38
     Protein
    -15.38
    鬘
    -15.31
     ref
    -15.19
    okes
    -15.00
     sauces
    -14.38
    POSITIVE LOGITS
     digit
    84.50
    数字
    82.50
     digital
    79.50
    digit
    79.50
     Digit
    77.50
    Digit
    75.00
    Digital
    73.50
     Digital
    73.50
    digital
    73.00
     digits
    71.00
    Activations Density 0.476%

    No Known Activations