© 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. 9-GEMMASCOPE-RES-16K
    6. 11412
    Prev
    Next
    INDEX
    Explanations

    mathematical and logical symbols, particularly in relation to comparisons and equations

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

    The neuron detects conjunctions, particularly "and" or semicolons, that separate items in a list or connect clauses.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 9-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_9/width_16k/average_l0_100
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.9.hook_resid_post
    Hook Layer
    9
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Franken
    -0.43
     Klo
    -0.43
     Bren
    -0.40
    ooth
    -0.39
     Wein
    -0.38
    Coupe
    -0.38
    <eos>
    -0.37
     Clements
    -0.35
     helix
    -0.35
     hos
    -0.34
    POSITIVE LOGITS
     defaultstate
    0.88
    WriteBarrier
    0.65
     Chwiliwch
    0.63
    Cyfeiriadau
    0.56
     lenker
    0.55
    +;
    0.54
     getArguments
    0.54
    giers
    0.54
    verwijspagina
    0.52
    ]};
    0.52
    Activations Density 0.393%

    No Known Activations