© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-2B
    5. 3-GEMMASCOPE-TRANSCODER-16K
    7. 3623
    Prev
    Next
    INDEX
    Explanations

    the word "hurt" and words with negative connotations

    oai_token-act-pair · gemini-2.0-flash

    hurt

    np_max-act · gemini-2.0-flash
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2b-pt-transcoders/layer_3/width_16k/average_l0_54
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.3.ln2.hook_normalized
    Architecture
    jumprelu_transcoder
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Negative Logits
     hurt
    -2.23
    hurt
    -1.59
     hurts
    -1.50
     Hurt
    -1.38
     injured
    -1.38
    <bos>
    -1.37
    Hurt
    -1.34
     hurting
    -1.32
     harmed
    -1.23
     harm
    -1.19
    POSITIVE LOGITS
    Abitanti
    0.86
     betweenstory
    0.84
     lenker
    0.80
    parsedMessage
    0.73
    PreferredItem
    0.72
    UnusedPrivate
    0.71
     TabLayout
    0.70
    AnchorStyles
    0.68
     transférez
    0.67
     externi
    0.66
    Activations Density 1.826%

    No Known Activations