© 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. 37-GEMMASCOPE-2-TRANSCODER-262K
    4. 123985
    Prev
    Next
    INDEX
    Explanations

    The explanation should be general enough to cover the diverse **TOP_POSITIVE_LOGITS** while still being specific to the **MAX_ACTIVATING_TOKENS** and **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**.The core theme in `MAX_ACTIVATING_TOKENS` is "score" and "pouring".The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` suggests context for scoring, like "very few runs", "score ability", "score relatively easily".The `TOP_POSITIVE_LOGITS` are diverse: `bags`, `files`, `நிதி`, `split`, `wires`, `Files`, `wires`, `wired`, `팜`, `scissor`.Some of these might relate to results or actions.scoring or actions

    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
    াহিনী
    0.39
    ندان
    0.38
     mq
    0.38
     വളരെ
    0.36
    的很
    0.36
    ﮨ
    0.36
    sterdam
    0.36
    গ্রামী
    0.36
     пережи
    0.35
    aternal
    0.35
    POSITIVE LOGITS
     bags
    0.47
    files
    0.45
     நிதி
    0.40
    split
    0.40
    wires
    0.40
    Files
    0.39
     wires
    0.39
    wired
    0.39
    팜
    0.39
     scissor
    0.38
    Activations Density 0.000%

    No Known Activations