© 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. Qwen3.5-4B
    3. 15-RES-MATRYOSHKA-65K
    4. 21546
    Prev
    Next
    INDEX
    Explanations

    The most relevant pattern observed is the occurrence of "JavaScript" followed by terms related to its status or context (like "enabled" or "'s"). The presence of "HTML" also strongly points to web development. The exotic characters and specific medical/technical terms in the `TOP_POSITIVE_LOGITS` and `TOP_ACTIVATING_TEXTS` suggest the neuron might be sensitive to technical jargon,

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    decoderesearch/qwen-3.5-saes/qwen-3.5-4b
    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
    端çĿĢ
    -0.05
    以ä¸Ĭå°±æĺ¯
    -0.05
    ero
    -0.05
     sahib
    -0.05
     pref
    -0.05
     rej
    -0.05
     reda
    -0.05
    orman
    -0.04
     clean
    -0.04
    åĿ¡
    -0.04
    POSITIVE LOGITS
    ount
    0.06
    angkit
    0.06
     asym
    0.06
    æ´»
    0.06
    æ´»çļĦ
    0.06
    天çĦ¶
    0.06
    沿
    0.06
    åIJįåī¯
    0.06
    ädt
    0.06
    ä¸ĵ
    0.06
    Activations Density 0.005%

    No Known Activations