© 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. 34-GEMMASCOPE-RES-131K
    7. 2915
    Prev
    Next
    INDEX
    Explanations

    remix, change, replace, removed, adjusted, installing, adding

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

    The neuron strongly activates on auxiliary verbs and past‐participle constructions (e.g. “had done,” “was created,” “were stolen”), effectively detecting passive or perfect verb phrases.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
     flexibilidad
    -1.05
     absorción
    -1.04
     refurbishment
    -1.03
     незавер
    -1.02
     berguna
    -1.00
    aliere
    -0.99
     reservar
    -0.97
     postponement
    -0.96
     cancelación
    -0.94
     preparado
    -0.94
    POSITIVE LOGITS
     change
    2.64
    เปลี่ยน
    2.22
     changes
    2.20
     changing
    2.16
     changed
    2.14
     remove
    2.03
    change
    1.97
     replace
    1.92
    changing
    1.84
     Change
    1.76
    Activations Density 0.163%

    No Known Activations