© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 22-GEMMASCOPE-2-TRANSCODER-262K
    4. 261530
    Prev
    Next
    INDEX
    Explanations

    The neuron detects determiners ("the", "an") followed by descriptive nouns or the end of text.Let's refine this based on the rules:- Concise (3-20 words).- Phrase, not a full sentence.- Find patterns.- Avoid "tokens", "patterns", "detects/predicts".- Specific.Looking again:MAX_ACTIVATING_TOKENS: Most common are 'the' (6/10), 'an' (2/10).TOKENS_AFTER_MAX_ACTIVATING_TOKEN: Most common is 'image' (5/10).determiners followed by image

    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_22_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
    ului
    0.18
     минус
    0.18
    貉
    0.18
    endlich
    0.18
     quadrado
    0.17
    相當
    0.17
     वगैरह
    0.17
     એક
    0.17
    ulho
    0.17
     attribu
    0.16
    POSITIVE LOGITS
    thro
    0.20
     લોકોને
    0.19
    频
    0.19
    addAction
    0.19
    ouncing
    0.18
    Increasing
    0.17
    စေ
    0.17
     aton
    0.17
     anxiety
    0.17
    боры
    0.17
    Activations Density 0.003%

    No Known Activations