© 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-3-12B-IT
    3. 12-GEMMASCOPE-2-RES-16K
    4. 9459
    Prev
    Next
    INDEX
    Explanations

    hookstobe,`test score 1` is placed *directly*testalgunameoperateitaddedLooking at these lists:* **MAX_ACTIVATING_TOKENS**: `to` appears twice, `hook` appears. `directly`, `,`, `hay`, `have`, `and`, `has` are also present.* **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `be` (after 'to'), `'` (after `hook`), `*` (after `,`), `Test` (after `directly`), `alguna` (after `hay`), `me` (after `have`), `operate` (after `have`), `it` (after `has`), `added` (after `has`).* **TOP_POSITIVE_LOGITS**: `blue`, `bluish`, `Blue`, `Run`, `Ruler`. These seem less directly related to the token patterns. `د` and `嶷` are likely non-Latin characters or symbols.* **TOP_ACTIVATING_TEXTS**: * "Add the hook to the window-selection-change and buffer-list-update hooks." - includes "hook" and "to" followed by "the". * "`//test score 1` is placed *directly* after the variable declaration." - includes "directly" followed by "after". * "What you *can* test, categorized by complexity:" - includes "test". * "10 hay un punto donde..." - includes "hay". * "...send compressed data to me (the LLM) and have me decompress it." - includes "have me".havetome

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-it/resid_post/layer_12_width_16k_l0_medium
    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
    ajout
    0.96
    必要がある
    0.88
    ïdes
    0.88
    }$}
    0.87
    zoek
    0.86
     besoins
    0.86
     utilises
    0.85
     amoureux
    0.84
    oyens
    0.84
    垐
    0.84
    POSITIVE LOGITS
     blue
    0.96
    د
    0.89
     bluish
    0.87
     Blue
    0.86
    k
    0.81
     Run
    0.77
     sloppy
    0.77
     Ruler
    0.76
     Nice
    0.75
    嶷
    0.75
    Activations Density 0.002%

    No Known Activations