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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 31-GEMMASCOPE-2-RES-65K
    4. 13901
    Prev
    Next
    INDEX
    Explanations

    debugging, surveillance, trafficking, political figures

    np_acts-logits-general · gemini-2.5-flash-lite

    # Neuron 4 Explanation This neuron detects **economically significant or resource-related keywords and phrases**. The neuron activates strongly on words like "caffeine," "prices tripled," "surveillance," "human trafficking," "programming," "cuddly," and technical specifications—terms that indicate important practical concerns about resources, costs, efficiency, risks, or technical optimization that the model emphasizes in its explanations.

    oai_token-act-pair · claude-4-5-haikuTriggered by @jamesnaruto04
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/resid_post/layer_31_width_65k_l0_medium
    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
    但是在
    0.52
    但在
    0.47
     แต่
    0.47
     लेकिन
    0.44
     but
    0.44
    但是我
    0.43
     אבל
    0.42
     있지만
    0.41
     flickering
    0.41
    TakePhoto
    0.41
    POSITIVE LOGITS
     Biden
    0.55
    Biden
    0.54
     Putin
    0.45
    āta
    0.45
     pretext
    0.42
     barrio
    0.42
     Bolsonaro
    0.42
     هذا
    0.42
     Bolshevik
    0.41
    Putin
    0.41
    Activations Density 0.002%

    No Known Activations