© 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. 17950
    Prev
    Next
    INDEX
    Explanations

    * "New York City, for plaintiffs."* "Pittsburgh, Pa., for plaintiff."* "Raleigh, for the State."The neuron detects commas followed by location names or organizational identifiers.**Phrase candidates based on pattern:**- commas before locations- comma followed by city/state- comma after punctuation- comma followed by legal entityLet's evaluate:- "commas before locations": This is good.- "comma followed by city/state": Also good, more specific.- "comma after punctuation": Too general.comma followed by location or entity

    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
     dynamics
    -0.06
     dõi
    -0.05
     Segment
    -0.05
    esper
    -0.05
     segment
    -0.05
     segmento
    -0.05
     specially
    -0.05
    loom
    -0.05
     пÑĢоÑĤÑı
    -0.05
    å¤ļäºİ
    -0.05
    POSITIVE LOGITS
    æīİæīİå®ŀ
    0.06
    ä¹Ŀä¹Ŀ
    0.06
    undi
    0.06
    reas
    0.05
    èĢĥçļĦ
    0.05
    kaan
    0.05
    éĻĭ
    0.05
    åľ°æĸ¹çļĦ
    0.05
    çŃīé¢Ĩ导
    0.05
    ãģ¾ãģĪ
    0.05
    Activations Density 0.005%

    No Known Activations