© 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. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 1-RES-SM
    11. 29141
    Prev
    Next
    INDEX
    Explanations

    repeated instances of the contraction "you're" in various contexts

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    ctigges/pythia-70m-deduped__res-sm_processed/1-res-sm
    Prompts (Dashboard)
    32,768 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.1.hook_resid_post
    Hook Layer
    1
    Architecture
    standard
    Context Size
    128
    Dataset
    EleutherAI/the_pile_deduplicated
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    478
    +0.22
    1.2%
    23
    +0.14
    0.8%
    371
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    371
    +0.22
    0.06
    348
    +0.14
    0.06
    473
    +0.11
    0.04
    Negative Logits
    Ĵ
    -2.06
    iliary
    -1.90
    º
    -1.86
    ļ
    -1.80
    ł
    -1.78
    OMA
    -1.78
    anto
    -1.77
    Ļª
    -1.71
    »¿
    -1.70
    ĸ
    -1.69
    POSITIVE LOGITS
     gonna
    1.95
     kidding
    1.81
     okay
    1.76
    wolf
    1.67
     ok
    1.58
    bent
    1.52
     going
    1.48
     joking
    1.48
    ptin
    1.47
    edy
    1.45
    Activations Density 0.243%

    No Known Activations