© 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. 3886
    Prev
    Next
    INDEX
    Explanations

    references to specific questions or requests for information

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

    punctuation marks and special characters, particularly periods and question marks at the end of sentences or statements.

    oai_token-act-pair · claude-4-5-haikuTriggered by @emiglarou

    line breaks or paragraph boundaries, particularly those that separate distinct sections of text or mark transitions between different topics or contexts.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    meta-text and structural markers of online posts—questions/exclamations, comment/Q&A prompts, citation brackets/numbers, dates/years, and “continue/return” headers—rather than regular prose content.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408
    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
    ]")]
    -0.61
    findpost
    -0.58
     NSCoder
    -0.58
     propOrder
    -0.56
    +#+
    -0.55
    Parcelize
    -0.54
     AssemblyCompany
    -0.54
    MessageTagHelper
    -0.50
     AssemblyProduct
    -0.49
     TextInputType
    -0.49
    POSITIVE LOGITS
    utilisons
    0.44
    épend
    0.43
    richTextPanel
    0.40
    쿵
    0.39
    itinéraire
    0.39
    IANGLES
    0.39
     sekali
    0.39
     habrá
    0.39
    исленность
    0.39
     Again
    0.39
    Activations Density 0.450%

    No Known Activations