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. 835
    Prev
    Next
    INDEX
    Explanations

    instances of the word "somehow" and its various forms

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

    expressions that imply uncertainty or unexpected outcomes

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

    somehow or for some reason

    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
    ⓧ
    -0.64
     ""));
    -0.60
    どうなる
    -0.53
    "]}
    -0.53
     XNUMX
    -0.52
     '.';
    -0.51
    < 
    -0.50
    "])){
    -0.50
     Audiodateien
    -0.49
     kính
    -0.48
    POSITIVE LOGITS
    Somehow
    1.10
     Somehow
    0.98
     magically
    0.95
     miraculously
    0.87
     somehow
    0.74
    culously
    0.71
    erweise
    0.68
    ParallelGroup
    0.67
     образом
    0.67
    findpost
    0.66
    Activations Density 0.028%

    No Known Activations