© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 37-GEMMASCOPE-2-TRANSCODER-262K
    4. 205284
    Prev
    Next
    INDEX
    Explanations

    The neuron seems to be detecting the end of a sentence or a significant break, often followed by specific types of tokens like 'Tools', 'K', 'We', 'G', or punctuation.Let's refine this by looking for a common theme.**MAX_ACTIVATING_TOKENS**: '.' is the only token provided. This is a punctuation mark.**TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: 'Tools', 'K', 'We', 'G', '('.The neuron detects sentence endings and subsequent context

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_37_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
     说
    0.46
    confirmation
    0.44
     uglify
    0.43
     mengubah
    0.42
     berakhir
    0.42
     EditorGUILayout
    0.41
     setError
    0.41
    ᒧ
    0.41
     terão
    0.40
    शियन
    0.40
    POSITIVE LOGITS
     هناك
    0.36
     AVI
    0.35
     Rot
    0.34
    ##
    0.34
    erical
    0.34
    чек
    0.34
     there
    0.33
     #
    0.33
    sce
    0.33
     instance
    0.33
    Activations Density 0.009%

    No Known Activations