© 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-3-1B
    3. 7-GEMMASCOPE-2-RES-16K
    4. 12304
    Prev
    Next
    INDEX
    Explanations

    也可能**Reasoning:**The most frequent token in `MAX_ACTIVATING_TOKENS` is "也" (also). It appears 7 out of 10 times.The most frequent token in `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` is "可能" (possibly/maybe), appearing 4 out of 10 times.The combination "也可能" (also possibly/may also) appears in several of the `TOP_ACTIVATING_TEXTS`, indicating a pattern of introducing possibilities, limitations, or alternative outcomes, often following a statement or condition. Examples include: "也可能导致他失去自我", "可能需要更多 VRAM", "也可能导致信息不一致"."但这" (but) also appears, often preceding a contrasting possibility or limitation. "esta frase" and "entire training" are less consistent across the lists."也可能" succinctly captures this frequent conjunction of possibility and alternative introduced by the neuron's activation, fitting the length requirement

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-1b-pt/resid_post/layer_7_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    vila
    2.19
    stehen
    2.11
    WithType
    2.06
    ('.')[
    2.05
    ℤ
    2.03
     einges
    1.95
    nst
    1.93
    bü
    1.93
    ből
    1.88
    nd
    1.87
    POSITIVE LOGITS
    જરા
    2.66
    γμα
    2.55
    ミック
    2.38
    ﻟ
    2.29
    侑
    2.26
    )){
    2.22
    ريق
    2.22
    逊
    2.20
    迟
    2.18
    olol
    2.17
    Activations Density 0.000%

    No Known Activations