Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-2B
    4. Residual Stream - 65k
    5. 20-GEMMASCOPE-RES-65K
    6. 1010
    Prev
    Next
    INDEX
    Explanations

    patterns or sequences within text that relate to HTML parsing and regex matching processes

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    elements and patterns related to regular expressions (regex)

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    regex patterns matched

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 20-gemmascope-res-65k
    Configuration
    google/gemma-scope-2b-pt-res/layer_20/width_65k/average_l0_114
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    65,536
    Data Type
    float32
    Hook Name
    blocks.20.hook_resid_post
    Hook Layer
    20
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    BeginInit
    -0.76
     <>",
    -0.69
    исленность
    -0.63
    WireFormatLite
    -0.60
    providedIn
    -0.58
    спонд
    -0.57
     Certi
    -0.53
     (!__
    -0.53
    󠁴
    -0.53
     easel
    -0.53
    POSITIVE LOGITS
     pattern
    1.46
     Pattern
    1.43
    Pattern
    1.42
     regex
    1.38
     patterns
    1.32
    pattern
    1.31
    regex
    1.30
     match
    1.26
     Regex
    1.25
     Match
    1.24
    Activations Density 0.628%

    No Known Activations