© 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-3-27B-IT
    3. 27-GEMMASCOPE-2-TRANSCODER-262K
    4. 196683
    Prev
    Next
    INDEX
    Explanations

    The user wants me to explain the neuron's behavior in a concise phrase (3-20 words) by finding patterns in the provided lists.Let's analyze the lists:1. **MAX_ACTIVATING_TOKENS**: `longer`, `longer`, `longer`, `longer`, `longer`, `longer`, `longer`, `longer`, `longer`, `longer` * This list is extremely repetitive, suggesting the neuron strongly activates when the word "longer" appears.2. **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `just`, `a`, `a`, `a`, `a`, `a`, `content`, `a`, `a`, `a` * Most of the tokens here are `a`. `just` and `content` appear less frequently.no longer something

    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_27_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
     internet
    0.33
     wheeled
    0.32
     container
    0.32
     abo
    0.31
     struck
    0.31
     Stanton
    0.31
     now
    0.31
     tapped
    0.31
     desalination
    0.31
     Internet
    0.30
    POSITIVE LOGITS
    höhe
    0.33
     sécrétion
    0.31
     مقابلہ
    0.29
    Vendo
    0.29
     patria
    0.29
    Ph
    0.28
    Appearance
    0.28
    巍
    0.28
     Haga
    0.28
     ทด
    0.28
    Activations Density 0.002%

    No Known Activations