© 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. 31-GEMMASCOPE-RES-16K
    6. 7703
    Prev
    Next
    INDEX
    Explanations

    terms related to cracks and fractures in materials

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

    mentions of cracks or fractures in physical or metaphorical contexts.

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

    words related to breaks, fractures, or structural damage.

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

    words describing physical breaks, splits, or structural damage (like "crack," "fracture," "flake").

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

    language about cracks and fractures—detecting the emergence, propagation, or signs of structural breakage in physical systems or metaphorical rifts in groups/support.

    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
    AnimationsModule
    -0.52
     informée
    -0.50
    Personendaten
    -0.46
    stylers
    -0.45
    IsPostBack
    -0.44
    ندران
    -0.43
    StoryboardSegue
    -0.42
     Dillon
    -0.42
    AntiForgeryToken
    -0.41
     enfans
    -0.41
    POSITIVE LOGITS
     cracks
    2.02
     crack
    1.81
     Cracks
    1.76
     cracked
    1.63
     cracking
    1.52
     Crack
    1.51
    crack
    1.48
     CRACK
    1.41
    Crack
    1.41
     fissures
    1.35
    Activations Density 0.309%

    No Known Activations