© 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. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 27995
    Prev
    Next
    INDEX
    Explanations

    option (a)

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

    The neuron fires strongly on question‐mark tokens (i.e. it detects question punctuation).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    multiple-choice answer options, particularly the letter identifiers (like "a)", "b)", "A.", "B.") that precede answer choices in test questions.

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

    multiple-choice question formatting, including question prompts and labeled answer options.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     what
    -1.16
     and
    -1.16
     or
    -1.13
     which
    -1.12
     that
    -1.09
     However
    -1.05
    ناصر
    -1.04
    อร่อย
    -1.02
     Although
    -1.02
    dö
    -1.02
    POSITIVE LOGITS
    /
    1.10
    beo
    1.06
    none
    1.04
     Dinas
    1.03
     colossal
    1.02
     cercanos
    0.96
    わせて
    0.94
    un
    0.94
    they
    0.93
     gigantic
    0.93
    Activations Density 0.011%

    No Known Activations