© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 15-GEMMASCOPE-2-TRANSCODER-262K
    4. 14649
    Prev
    Next
    INDEX
    Explanations

    "code syntax" directly addresses what the neuron is picking up on. The presence of symbols like `<`, `>`, `"` in `MAX_ACTIVATING_TOKENS`, followed by "module" and "line" in `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`, points towards code-related tokens. The `TOP_POSITIVE_LOGITS` showing things like `):`, `);`, `Oops`, `");` further reinforces the idea of detecting syntax, potentially for code completion or error detection. This phrase is concise and specific.* **Conciseness:** 2 words.* **Pattern:** Detects common programming syntax elements.* **Avoided:** "

    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_15_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
     REPORTS
    0.43
     প্রতিভা
    0.43
     With
    0.42
     annuities
    0.42
     embora
    0.42
     talento
    0.41
     anuncios
    0.41
     Presumably
    0.40
     verlie
    0.40
     굉장
    0.40
    POSITIVE LOGITS
     ):
    0.52
    oc
    0.50
     );
    0.50
     ओ
    0.46
    o
    0.46
    Oops
    0.46
     ");
    0.43
    𝗼
    0.43
     o
    0.43
    โอ
    0.42
    Activations Density 0.006%

    No Known Activations