© 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
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 33428
    Prev
    Next
    INDEX
    Explanations

    contrast and conditional outcomes

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

    The neuron is highly sensitive to contrastive discourse markers (e.g. “however,” “but”) that signal an adversative or shifting relation in the text.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     monstre
    -0.93
    already
    -0.91
     Lave
    -0.89
    何より
    -0.86
     zaten
    -0.82
    iog
    -0.82
     marcher
    -0.82
    க்கும்
    -0.81
     überhaupt
    -0.81
     giang
    -0.79
    POSITIVE LOGITS
     should
    1.65
     deveria
    1.32
     SHOULD
    1.29
     would
    1.27
     could
    1.23
     Should
    1.23
     ought
    1.20
     someday
    1.11
     debería
    1.09
    should
    1.08
    Activations Density 0.077%

    No Known Activations