© 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
    3. Gemma-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 9268
    Prev
    Next
    INDEX
    Explanations

    Roman numerals and specific punctuation

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

    The neuron detects document structure markers—especially Roman‐numeral section or figure labels (e.g. “III.”, “II-1”, “VII-99”) and related numbering patterns.

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

    references and labels that use Roman numerals, especially when combined with dashes/periods/parentheses in section titles, figure numbers, ranges, and coded designations.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408

    Roman numerals and section markers in technical and formal documents.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/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
    '
    -2.69
    0
    -2.44
    D
    -2.27
    :
    -2.25
    ”
    -2.03
    ↵↵
    -2.00
    —
    -1.85
    R
    -1.84
    -
    -1.82
    7
    -1.78
    POSITIVE LOGITS
    ﻒ
    2.06
     emprender
    1.95
    
    1.80
    泐
    1.80
    ):}
    1.80
     desbloquear
    1.80
    ྵ
    1.80
     encaje
    1.79
    вольно
    1.76
    鹇
    1.75
    Activations Density 0.048%

    No Known Activations