© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 16-GEMMASCOPE-2-TRANSCODER-262K
    4. 216134
    Prev
    Next
    INDEX
    Explanations

    **Explaining Neuron Activity**I'm analyzing the given lists to pinpoint the core function of this neuron. I'll examine the sequence of tokens and the context in which they appear to distill its behavior into a concise phrase.The provided lists show repeated instances of:* `registration` followed by `permit`* `Merge`, `merge`, `merges` followed by `onto` or `into`* `ordering` followed by context implying choices (self-service)* `service` (counter service, delivery services)* `roundabout` (navigation)These all relate to actions associated with gaining access, entering a new state or location, or initiating a process of acquisition. The keywords suggest instructions, processes, or transitions.A concise phrase capturing this would be related to entering or obtaining.Let's try to find the best fit:- "entering and obtaining" - a bit too general, feels like a description.- "navigation and transactions" - good, but maybe `registration` leans more towards permission than transaction.- "initiate entry or acquire" - getting closer to action.- "merging, ordering, and service" - too specific to list them all.Consider the actions they describe:- `registration permit`: acquiring permission to enter.- `Merge onto highway`: entering a traffic flow.- `ordering`: initiating a request for something.- `service`: the act of providing/getting something.- `roundabout`: navigating a transition point.The core idea is *transitioning into* something or *acquiring* something."transitioning into or acquiring" is 4 words."entry and acquisition initiating" is 4 words.Let's look at the specific tokens `Merge` and `ordering`. The outputs are `onto` and `into`.`registration` -> `permit`.entering or acquiring

    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_16_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
    AH
    0.50
     F
    0.49
     D
    0.49
    <i>
    0.47
    ERROR
    0.46
    Nota
    0.46
    Ah
    0.46
    De
    0.45
    Note
    0.45
     De
    0.44
    POSITIVE LOGITS
    솎
    0.57
     બનાવવા
    0.55
    两国
    0.54
    hips
    0.53
    হইবে
    0.53
     Janata
    0.52
    сиона
    0.52
     රට
    0.51
     ситуа
    0.51
    ijala
    0.51
    Activations Density 0.014%

    No Known Activations