Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-2B
    4. Residual Stream - 16k
    5. 18-GEMMASCOPE-RES-16K
    6. 15327
    Prev
    Next
    INDEX
    Explanations

    sentences related to social interactions and personal feelings

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

    after certain question-related tokens

    np_acts-logits-general · gemini-2.0-flash

    index array

    np_acts-logits-general · gemini-2.5-flash-lite

    Finds mentions of bugs/problems/glitches or issue reports, often with accompanying time durations or status (e.g., months, reported/fixed, version, translation).

    oai_token-act-pair · gpt-5-miniTriggered by @igorkuztech

    pronouns and commas Method used: 3 — top logits are pronouns/punctuation, so "pronouns and commas"

    np_max-act-logits · gpt-5-miniTriggered by @igorkuztech
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 18-gemmascope-res-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_18/width_16k/average_l0_74
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.18.hook_resid_post
    Hook Layer
    18
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    )’
    -0.77
     }}$}
    -0.73
    .」
    -0.71
    ]."
    -0.70
    ’.”
    -0.69
     дописавши
    -0.68
    .")
    
    -0.68
    ?")
    -0.67
    ."]
    -0.67
    ).”
    -0.67
    POSITIVE LOGITS
     ,
    0.83
     it
    0.82
     I
    0.78
     please
    0.73
     we
    0.71
     they
    0.71
     It
    0.66
     They
    0.65
     semoga
    0.61
     but
    0.61
    Activations Density 0.259%

    No Known Activations