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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & 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. 20-GEMMASCOPE-RES-16K
    6. 2597
    Prev
    Next
    INDEX
    Explanations

    conversations or official communications regarding government or emergency responses.

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

    statements related to government meetings, legislative processes, and discussions on establishing laws.

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot

    This neuron primarily activates for verbs of speech reporting, particularly the word "told" followed by a subject (often a news outlet or reporter). It seems to be looking for phrases that indicate a quote or statement is being attributed to a source, especially in a journalistic context.

    oai_token-act-pair · claude-3-5-sonnet-20240620Triggered by @bot

    Told, followed by a reporter/news agency

    np_acts-logits-general · gemini-2.0-flash

    told reporters

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.20.hook_resid_post
    Hook Layer
    20
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     للاسماء
    -0.69
     estekak
    -0.63
     femininas
    -0.60
     devriez
    -0.54
    uxxxx
    -0.53
     sauvages
    -0.53
     particulières
    -0.53
    resave
    -0.52
     positifs
    -0.52
     élevées
    -0.52
    POSITIVE LOGITS
     reporters
    1.27
     reporter
    0.89
     journalists
    0.88
     interviewer
    0.86
     periodistas
    0.80
     interviewers
    0.80
    记者
    0.74
     investigators
    0.73
    الحياه
    0.69
     wartawan
    0.69
    Activations Density 0.092%

    No Known Activations