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. Gemma-3-4B-IT
    3. 11-GEMMASCOPE-2-TRANSCODER-262K
    4. 244522
    Prev
    Next
    INDEX
    Explanations

    The neuron fires strongly on numeric tokens—digits, numbers (including years), and related symbols (like “=”)—i.e. it’s looking for numerals/math expressions.

    oai_token-act-pair · o4-miniTriggered by @drejcpesjakpesjak

    interrogative, quiz-style questions—especially sentences with question formatting, wh-words, and simple fact or math queries.

    oai_token-act-pair · gpt-5Triggered by @drejcpesjakpesjak
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-4b-it/transcoder_all/layer_11_width_262k_l0_small_affine
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     karena
    1.03
     Exclusion
    1.01
     depletion
    0.98
     exclusions
    0.98
     dismissal
    0.93
     expresiones
    0.92
     Dried
    0.91
     pengurangan
    0.90
     condo
    0.90
    ক্লো
    0.90
    POSITIVE LOGITS
    М
    0.75
    Java
    0.70
    ('
    0.69
    continued
    0.69
    Leet
    0.69
    SEL
    0.69
    perf
    0.69
    abbr
    0.67
    spell
    0.67
    Perf
    0.66
    Activations Density 0.103%

    No Known Activations