© 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. 32-GEMMASCOPE-2-TRANSCODER-262K
    4. 234559
    Prev
    Next
    INDEX
    Explanations

    The neuron is activated by the keyword `while` followed by conditional statements or loops that involve counting, length checks, or prime number calculations.Looking at `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`, we see `(`, `count`, `len`, `prime`, `(!`.Looking at `TOP_ACTIVATING_TEXTS`, we see many examples of `while` followed by:* `count <`* `len(`* `prime` related constructs (although `prime` itself isn't directly after `while` in the examples, it's in the text related to the `while` loop)* `(!` (for boolean checks)* `True`The `TOP_POSITIVE_LOGITS` seem disparate and largely irrelevant here, as the pattern is clearly in the structure of the code snippets.The most consistent pattern is `while` followed by a condition, often related to iterative processes like counting, checking length, or reaching a target.Possible explanations:* `while` loops condition* looping conditions* control flow statements `while`* iterative constructs `while`* `while` loop context* `while` and loop conditionsLet's refine this. The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` and `TOP_ACTIVATING_TEXTS` strongly point towards `while` loops and their conditions."loop conditions" is 2 words."while loop conditions" is 3 words."iterative loop conditions" is 3 words.The neuron seems to detect the 'while' keyword in the context of initiating or controlling loops based on certain conditions.The `TOP_POSITIVE_LOGITS` are a red herring here. The action of `while` followed by a conditional check is the strong pattern.Consider the phrase: "while loop conditions" (3 words)Consider the phrase: "iterative control flow" (3 words)Consider the phrase: "loop initiation conditions" (3 words)The key elements are `while`, and then a condition that allows loops.while loop conditions

    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_32_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
     rental
    0.41
     taf
    0.39
     روحانی
    0.39
     medical
    0.38
     කර
    0.37
     медицин
    0.37
     tourist
    0.37
    KCl
    0.36
     Navarre
    0.36
     required
    0.36
    POSITIVE LOGITS
    不住
    0.38
     அற்புதமான
    0.37
    inator
    0.36
    杀了
    0.36
     Monate
    0.35
    ევე
    0.35
     காலத்தில்
    0.35
     Ora
    0.35
    ือก
    0.34
    ছাড়া
    0.34
    Activations Density 0.001%

    No Known Activations