© 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. 24-GEMMASCOPE-2-TRANSCODER-262K
    4. 176373
    Prev
    Next
    INDEX
    Explanations

    analyzing text lists.**MAX_ACTIVATING_TOKENS**: `to`, `Diff`, `7`, `):`**TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `use`, `?`, `the`, `raction`, `located`, `the`, `0`, `images`Looking at the `TOP_POSITIVE_LOGITS` and `TOP_ACTIVATING_TEXTS` for context is crucial.**TOP_POSITIVE_LOGITS**: `orada`, `rinsim`, `થો`, `thoracic`, `ಾಯಿತು`, `desiderio`, `ormal`, `usa`, `ලය`, `standardized`These are multilingual (Hindi, Gujarati, Spanish, Turkish, English) and seem diverse, but `thoracic` and `standardized` might hint at more technical or descriptive language. `usa` is a common country code.**TOP_ACTIVATING_TEXTS**:- "...capability for grammer which means it can demonstrate the ability to combine words and be able to use grammatical constructions.Analyzing the provided lists:* **MAX_ACTIVATING_TOKENS**: `to`, `Diff`, `7`, `):`* **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `use`, `?`, `the`, `raction`, `located`, `the`, `0`, `images`The token `7` followed by `raction` strongly suggests the word "diffraction". Some texts mention particles behaving like waves and encountering obstacles. `TOP_POSITIVE_LOGITS` includes `thoracic`, which is a specific anatomical term.Given the limited information and potential for noise, focusing on the most striking sequence:"7" followed by "raction" points directly to "diffraction". Other items don't form a clear, consistent pattern across the board that can be generalized easily beyond this specific observation, especially with the multilingual aspect of `TOP_POSITIVE_LOGITS`.Therefore, the most specific detectable pattern is related to diffraction.diffraction

    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_24_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
    ający
    0.48
    联
    0.45
     பார
    0.45
    ř
    0.44
     వివ
    0.42
    ון
    0.41
     আলোচিত
    0.41
     দেয়া
    0.39
    ليك
    0.38
    änz
    0.38
    POSITIVE LOGITS
     orada
    0.48
    rinsim
    0.45
     થો
    0.45
     thoracic
    0.43
    ಾಯಿತು
    0.41
     desiderio
    0.41
    ormal
    0.40
    usa
    0.39
    ලය
    0.39
     standardized
    0.39
    Activations Density 0.000%

    No Known Activations