© 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. 21-GEMMASCOPE-TRANSCODER-16K
    7. 1806
    Prev
    Next
    INDEX
    Explanations

    references to studies or data

    oai_token-act-pair · gemini-2.0-flash

    past tense verbs starting with 'did'

    np_token-act-pair-logits · gpt-4o-mini

    past tense verbs

    np_max-act-logits · gemini-2.0-flash

    say "past-tense verbs" Method used: 2 — because following tokens are predominantly past/past-participial forms; say "past-tense verbs"

    np_max-act-logits · gpt-5-mini
    Triggered by @chenshw0109
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2b-pt-transcoders/layer_21/width_16k/average_l0_13
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.21.ln2.hook_normalized
    Architecture
    jumprelu_transcoder
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Negative Logits
     makes
    -0.66
     goes
    -0.64
     is
    -0.62
     does
    -0.61
     becomes
    -0.60
     wordt
    -0.59
     gives
    -0.59
     gets
    -0.57
     reduces
    -0.56
     takes
    -0.56
    POSITIVE LOGITS
     did
    1.29
     took
    1.22
     lasted
    1.18
     lacked
    1.16
     remained
    1.16
     consisted
    1.16
     didn
    1.14
     went
    1.13
     didnt
    1.13
     came
    1.11
    Activations Density 6.671%

    No Known Activations