© 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. 37-GEMMASCOPE-2-TRANSCODER-262K
    4. 205295
    Prev
    Next
    INDEX
    Explanations

    "focus on your" is the most concrete pattern observed from `MAX_ACTIVATING_TOKENS` (where "on" appears) and `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` (where "your" appears after "on"). This directly relates to the content of `TOP_ACTIVATING_TEXTS` which are heavily about meditation and focusing.The digits don't seem to form a clear pattern.The `TOP_POSITIVE_LOGITS` list (`"}`, `Nft`, `ClN`, `Won`, `Development`, `nft`, `DEVELOPMENT`, `MoveNext`, `δικ`, `LevelLinking`) is quite eclectic and doesn't strongly support or contradict a specific pattern emerging from the token lists for this particular neuron. The presence of "nft" and "Development" might suggest a topic area, but the more granular token analysis points to a structural or functional pattern.Given the prompt to find patterns in lists and explain the neuron's behavior, the recurring sequence "on your" is the most prominent and directly observable pattern. The `TOP_ACTIVATING_TEXTS` reinforce this, as phrases like "Focus on your breath", "focus on your body sensations", "focus on your breath", "focus on your breath" are abundant.Therefore, the behavior of the neuron is best described by what it seems to be looking for or activating when "on your" follows a primary activation point.Final Answer: **focus on your**focus on your

    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_37_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
     stress
    0.52
     estrés
    0.52
     Stress
    0.51
     aktiviert
    0.48
     stres
    0.45
     instância
    0.45
    Stress
    0.44
     instancias
    0.44
    ால்
    0.42
     démontré
    0.42
    POSITIVE LOGITS
    }");
    0.38
    Nft
    0.38
    ClN
    0.36
    Won
    0.36
    Development
    0.36
     nft
    0.36
     DEVELOPMENT
    0.34
    MoveNext
    0.34
    δικ
    0.34
    LevelLinking
    0.34
    Activations Density 0.007%

    No Known Activations