© 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. 7-GEMMASCOPE-2-TRANSCODER-262K
    4. 31912
    Prev
    Next
    INDEX
    Explanations

    The presence of various programming language syntax, especially `int`, `arc`, and assignments (`=`), suggests the neuron is keyed into code structures. The subsequent tokens like `n`, `g`, `canvas`, `item` further reinforce this. The mix in `TOP_POSITIVE_LOGITS` is less clear but could indicate broad language handling or noise.Considering the prompt asks for a pattern *finding* and giving a phrase about what it detects/predicts.1. **MAX_ACTIVATING_TOKENS**: Mostly symbols like `=`, `:`, `"` - these are very common in code for assignments, definitions, string literals.2. **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `n`, `Int`, `g`, `"`, `Arc`, `Pawn`, `3`, `canvas`, `item`. This is a mixed bag, but `Int`, `Arc` strongly point to programming. `Pawn` is a specific word, `canvas` graphics. `item` is generic.3.code syntax and structures

    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_7_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
     of
    0.64
    )،
    0.63
    鸹
    0.60
     unbalanced
    0.59
    '
    0.59
     to
    0.57
     be
    0.57
     冖
    0.55
     опла
    0.55
     власти
    0.55
    POSITIVE LOGITS
    ர்
    0.84
    v
    0.82
    ان
    0.78
    ра
    0.77
    ر
    0.77
    z
    0.75
    و
    0.73
    א
    0.70
    ம்
    0.69
    f
    0.68
    Activations Density 0.038%

    No Known Activations