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-2-9B
    4. Residual Stream - 131k
    5. 20-GEMMASCOPE-RES-131K
    6. 103022
    Prev
    Next
    INDEX
    Explanations

    occurences of the term '<bos>' indicating the beginning of a block or section in the text

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

    The token "<bos>" marks the beginning of a new text segment or document, acting as a delimiter between different content sources in training data.

    eleuther_acts_top20 · claude-3-7-sonnet-20250219Triggered by @lu-christina
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    131,072
    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
    帖最后由
    -0.75
    transQ
    -0.65
    PerformLayout
    -0.65
    PMailer
    -0.61
    AutoField
    -0.57
    AnchorTagHelper
    -0.57
    .*")]
    -0.56
     <<<<<<<<<<<<<<
    -0.56
    astify
    -0.56
     CanadaChoose
    -0.52
    POSITIVE LOGITS
    DD
    0.38
    Taken
    0.35
     Zjednoc
    0.34
    YO
    0.34
     NIS
    0.33
     improving
    0.33
    起
    0.33
    雀
    0.33
     Taken
    0.32
     sho
    0.32
    Activations Density 0.078%

    No Known Activations