© 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. 93066
    Prev
    Next
    INDEX
    Explanations

    unbelievably amazing

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

    The neuron fires on garbled or non-standard tokens—i.e. text segments that look like OCR/mis-scanned or otherwise corrupted strings rather than normal words.

    oai_token-act-pair · o4-miniTriggered 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
     through
    -3.17
     one
    -3.11
     before
    -3.09
     as
    -3.08
     after
    -3.06
     from
    -3.06
     because
    -2.98
     using
    -2.92
     what
    -2.91
     for
    -2.89
    POSITIVE LOGITS
     Dinas
    2.98
     !!!!
    2.84
    commonly
    2.78
     delightfully
    2.70
     😁
    2.64
     !!!!!
    2.59
     REALLY
    2.58
     unbelievably
    2.56
     ridiculously
    2.53
     astounding
    2.52
    Activations Density 0.009%

    No Known Activations