© 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. 17662
    Prev
    Next
    INDEX
    Explanations

    It seems like the user wants me to act as a system that explains neuron behavior.I need to output a phrase (3-20 words) describing the pattern.The phrase should be specific, avoid meta-commentary, and not start with "This neuron detects".Let's re-examine the lists, focusing on finding the strongest pattern.* **MAX_ACTIVATING_TOKENS**: `opt`, `time`, `in`, `Ryan`, `:`, `a`, `. ` * The most notable token here is `Ryan`.* **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `g`, `.word`, `=`, `Debug`, `<`, `writing`, `Reynolds`, `ï`, `.` * Crucially, after `Ryan` comes `Reynolds`.Ryan Reynolds

    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
    éī´
    -0.06
    rim
    -0.05
    opak
    -0.05
    åīį
    -0.05
    饼
    -0.05
     Sus
    -0.05
     prem
    -0.05
    ä¸ĥå¤ķ
    -0.05
    leck
    -0.05
    çīµ
    -0.05
    POSITIVE LOGITS
    )â̦
    0.06
    íĴĪìĿĦ
    0.06
     blows
    0.06
    DDL
    0.06
    zny
    0.06
    temps
    0.05
    angers
    0.05
     blow
    0.05
    mener
    0.05
    orno
    0.05
    Activations Density 0.005%

    No Known Activations