© 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
    3. Gemma-3-27B-IT
    5. 27-GEMMASCOPE-2-TRANSCODER-262K
    7. 117884
    Prev
    Next
    INDEX
    Explanations

    The user wants me to act as a neuron and explain my behavior.The explanation must be:- Concise (3 to 20 words).- A phrase, not a full sentence.- About what I detect or predict by finding patterns in lists.- Specific, not general like "unique words".- Avoids phrases like "This neuron detects/predicts", "tokens", "patterns".- No capitalization unless it's a proper noun.I have the following lists:- `MAX_ACTIVATING_TOKENS`: `hung`- `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`: `r`, `1`, `1`, `5`, `5`, `A`, `.`, `source`, `years`, `rac`- `TOP_POSITIVE_LOGITS`: `hoạch`, `fiscal`, `para`, `Dirt`, `Fiscal`, `විස`, `Plan`, `allgemein`, `dirt`, `Functional`- `TOP_ACTIVATING_TEXTS`: Contains texts about "hungrily approaching", "losses", "GameStop's 2021 Squeeze", "175th anniversary", "50-Seat Bonus", "50,000-year-old clone", "initial A-share listing", "1.5-year-old child", "ChatGPT open source models", "10 years of experience", "20 years in exile".Let's break down the patterns:1.words like fiscal, plan, dirt

    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
    ᱙
    0.38
    quee
    0.34
    ະລ
    0.33
    깽
    0.33
     accustomed
    0.32
    sede
    0.32
    urun
    0.31
    Ⅸ
    0.31
    ❞
    0.31
    𝒑
    0.31
    POSITIVE LOGITS
     hoạch
    0.37
     fiscal
    0.35
     para
    0.32
     Dirt
    0.32
     Fiscal
    0.31
     විස
    0.31
     Plan
    0.30
     allgemein
    0.30
     dirt
    0.29
     Functional
    0.29
    Activations Density 0.000%

    No Known Activations