© 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. 206211
    Prev
    Next
    INDEX
    Explanations

    The MAX_ACTIVATING_TOKENS list contains primarily punctuation marks like `;`, `.*`, and `.*`. The TOKENS_AFTER_MAX_ACTIVATING_TOKEN list is mostly empty. The TOP_ACTIVATING_TEXTS list contains code snippets from various programming languages (TypeScript, JSX, Kotlin/JVM, C#).The pattern seems to be that the neuron activates when it encounters specific punctuation characters (like `;` or `.*`) in code, especially when these characters mark the end of a line or are followed by nothing significant. The TOP_POSITIVE_LOGITS are less directly related to this punctuation pattern and might be a secondary activation, but the most striking pattern from MAX_ACTIVATING_TOKENS points to these specific characters in a code context.Considering the rules:- Concise (3-20 words)- Find a pattern- Could be about tokens following or preceding certain tokens.- Avoid listing all tokens.- "Just say the pattern itself"- No "This neuron detects/predicts", "tokens", "patterns".- Specific.The MAX_ACTIVATING_TOKENS are punctuation

    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
     sombre
    0.43
     visual
    0.42
     perishable
    0.39
     builders
    0.39
     soluble
    0.38
     اجيب
    0.38
     linear
    0.37
     miserable
    0.37
     architects
    0.37
     ape
    0.36
    POSITIVE LOGITS
     Update
    0.41
    Update
    0.38
     प्रश्नावली
    0.38
    Nev
    0.38
     Hourly
    0.37
    舸
    0.37
    passwd
    0.36
    QRCode
    0.36
    pd
    0.36
     Voir
    0.36
    Activations Density 0.000%

    No Known Activations