© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26
    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 27-GEMMASCOPE-2-TRANSCODER-262K
    4. 193264
    Prev
    Next
    INDEX
    Explanations

    The numbers are followed by colons, indicating time formats (e.g., `09:30`). The top positive logits include words like 'incarcerated', 'sidelined', 'clamped', 'abducted', and 'consumed', which often relate to a state or event affecting something or someone. The presence of dates and timestamps in the activating texts further reinforces the temporal aspect.Therefore, the neuron seems to be detecting or signaling states or events occurring at specific times, particularly when numbers are formatted as time.Explanation: **time formats and states**time formats and states

    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_27_width_262k_l0_small_affine
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Negative Logits
    磪
    0.43
     sigu
    0.39
     Latest
    0.38
     Vou
    0.37
    Latest
    0.36
    citealt
    0.35
     Ці
    0.35
    https
    0.35
     kiló
    0.35
     https
    0.34
    POSITIVE LOGITS
     incarcerated
    0.43
     sidelined
    0.43
    পাতালে
    0.41
     clamped
    0.40
     maxHeight
    0.40
    గాలు
    0.40
     abducted
    0.39
     adsorbed
    0.38
     ovarian
    0.38
     consumed
    0.38
    Activations Density 0.000%

    No Known Activations