© 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. 104351
    Prev
    Next
    INDEX
    Explanations

    contractions

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

    The neuron is picking up on document‐layout and transcript metadata tokens—things like stray punctuation, parentheses, timestamps, labels (e.g. “VIDEO,” “CLIP”), page or line numbers—rather than content 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
     This
    -1.21
    writeHead
    -1.08
     undersø
    -0.99
     Only
    -0.98
     only
    -0.96
     see
    -0.95
    桩
    -0.93
    ❉
    -0.92
     These
    -0.91
     menulis
    -0.91
    POSITIVE LOGITS
     and
    1.34
     pistolet
    1.05
     modrá
    1.05
     jambes
    1.03
     bolsillos
    1.02
     erhi
    1.02
    lastung
    1.01
    pulled
    1.01
     dormitorios
    1.00
    meditation
    1.00
    Activations Density 0.004%

    No Known Activations