© 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. 12-GEMMASCOPE-2-TRANSCODER-262K
    4. 137800
    Prev
    Next
    INDEX
    Explanations

    Based on the provided lists, the neuron seems to be associated with administrative actions and access control, particularly within organized contexts like software, servers, or communities.Here's a breakdown of how the lists point to this:* **MAX_ACTIVATING_TOKENS**: Keywords like "Permissions," "permission," "administration," and "administrators" are extremely strong indicators. "org" appearing multiple times suggests a context of organizations or communities. "Cheat" might refer to commands or game mechanics often managed by administrators.* **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: While mostly punctuation, the presence of "community" after some activating tokens hints at the sphere of influence for these permissions.* **TOP_POSITIVE_LOGITS**: Words like "System," "Website," "Architectural," and "Builds" point towards technical or structural contexts where permissions are crucial. "Future" could relate to planning or policy.The pattern is clearly about managing access and rules within structured environments.**Explanation:**administration, permissions, authorization

    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_12_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
     diplomats
    0.68
     withdrawn
    0.65
     obscene
    0.64
     cinereo
    0.60
     И
    0.59
     stoichi
    0.59
    ตอบ
    0.59
     τραγ
    0.58
    ᅵ
    0.58
    ために
    0.57
    POSITIVE LOGITS
     System
    0.58
     Website
    0.55
    ţia
    0.51
     MRT
    0.51
     Architectural
    0.50
     Future
    0.50
    莒
    0.50
     Builds
    0.48
     CMB
    0.48
    ád
    0.48
    Activations Density 0.001%

    No Known Activations