© 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-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 74965
    Prev
    Next
    INDEX
    Explanations

    not necessarily

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

    The neuron fires on words that introduce negation, qualification, or hedging—e.g. “not,” “might,” “may,” “while,” and similar contrastive or uncertain cue words.

    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
    それにしても
    -1.08
    そも
    -1.05
    BIDDEN
    -0.92
     télécommande
    -0.91
     hvit
    -0.91
     kahit
    -0.90
    𝟸
    -0.90
    他に
    -0.90
    そういえば
    -0.88
    けません
    -0.88
    POSITIVE LOGITS
     necessarily
    1.37
     perfect
    1.12
     конечно
    1.09
     but
    1.09
     может
    1.02
    urou
    1.02
     allerdings
    1.00
    可能
    0.98
     vielleicht
    0.96
    jarat
    0.95
    Activations Density 0.030%

    No Known Activations