© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-9B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 2994
    Prev
    Next
    INDEX
    Explanations

    words and phrases related to toggling, switching, or controlling states in various contexts

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

    binary/toggle states and switches, especially technical terms related to turning things on and off.

    oai_token-act-pair · claude-3-5-haiku-20241022Triggered by @neilrathi

    The neuron identifies binary states or opposites, particularly "on/off" and similar toggle concepts like "up/down" or "in/out" in technical contexts.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi

    phrases that express binary opposites or toggled states/directions, especially when paired with connectors like “and/or” or a slash.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408

    the word "off" in the context of toggling, switching, or binary states (on/off).

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_20/width_16k/average_l0_68
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    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
     mijne
    -0.47
     zijne
    -0.40
     pouvoit
    -0.37
     feroit
    -0.37
     enfans
    -0.36
     plufieurs
    -0.36
    botella
    -0.35
     frühen
    -0.34
     ELECTRON
    -0.34
    žený
    -0.33
    POSITIVE LOGITS
    AndEndTag
    0.76
    存于互联网档案馆
    0.73
    ChildScrollView
    0.70
    KommentareTeilen
    0.65
    TargetException
    0.64
    aarrggbb
    0.63
    EndContext
    0.63
    AnimationsModule
    0.63
    fillType
    0.60
    ymce
    0.60
    Activations Density 0.424%

    No Known Activations