© 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. 10-GEMMASCOPE-2-TRANSCODER-262K
    4. 48944
    Prev
    Next
    INDEX
    Explanations

    Kes. **TOP_ACTIVATING_TEXTS**: - "*Oi Frog! series by Kes Gray & Jim Field:*" - "Kes" appears. - "Kesh (Uncut Hair): Sikhs do not cut their hair or beard" - "Kesh" appears. - "('kesusilaan')." - "kes" appears (part of a longer word). - "Kimi to issho ni, mite yuku keshiki" - "keshiki" appears. - "Viselj kesztyűt" - "kesztyűt" appears. - "Kessoku Band." - "Kes" appears. - "gkesws." - "kes" appears (part of a longer word). - "Keszthely (Festetics Palace)." - "Kes" appears.**Analysis:**- The `MAX_ACTIVATING_TOKENS` are consistently variations of "kes" or "esh".- The `TOP_ACTIVATING_TEXTS` show a strong presence of words starting with or containing "kes". Examples include author names ("Kes Gray"), religious terms ("Kesh"), foreign words ("kesusilaan", "kesztyűt", "Kessoku", "Keszthely"), and Japanese words ("keshiki").- The `TOP_POSITIVE_LOGITS` are diverse, indicating the neuron might activate for many different kinds of words, possibly related to the initial trigger.- The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` are varied and don't offer a consistent following pattern, but the texts themselves provide the primary clue.The neuron appears to be strongly associated with the syllable "kes" or "Kesh", occurring in names, foreign words, and specific concepts.**Result:** "Kes"

    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_10_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
    การ
    0.87
    Sh
    0.84
     uang
    0.80
    {
    0.76
     it
    0.72
    ı
    0.71
    Μ
    0.70
     be
    0.69
     строку
    0.69
    大
    0.69
    POSITIVE LOGITS
    at
    0.91
     breezy
    0.81
    ຢູ່ໃນ
    0.79
    ामान्य
    0.78
    urés
    0.77
    elijke
    0.75
    risis
    0.73
    restore
    0.71
    flatable
    0.71
    риев
    0.71
    Activations Density 0.001%

    No Known Activations