© 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
    3. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    5. Gemma-2-9B
    7. Residual Stream - 16k
    9. 31-GEMMASCOPE-RES-16K
    11. 13063
    Prev
    Next
    INDEX
    Explanations

    phrases related to scientific research and methodology

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

    This neuron identifies commas that separate items in a list or that follow technical terms, especially in scientific or academic writing.

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

    commas and periods that mark the end of clauses or sentences in technical and scientific writing.

    oai_token-act-pair · claude-4-5-haikuTriggered by @emiglarou
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 31-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_31/width_16k/average_l0_114
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.31.hook_resid_post
    Hook Layer
    31
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    bootstrapcdn
    -0.70
     resourceCulture
    -0.45
    mybatisplus
    -0.43
     defStyleAttr
    -0.43
    bufio
    -0.42
     виправивши
    -0.40
    adell
    -0.39
    ветить
    -0.39
    lloworld
    -0.38
    reactivex
    -0.38
    POSITIVE LOGITS
     while
    3.42
     whereas
    3.22
     mentre
    3.03
    while
    2.95
     mientras
    2.94
     tandis
    2.89
    whereas
    2.81
     sedangkan
    2.75
     enquanto
    2.69
     Whereas
    2.58
    Activations Density 1.834%

    No Known Activations