© 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. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 2-RES-SM
    11. 32381
    Prev
    Next
    INDEX
    Explanations

    instances of the word "of" in various contexts, indicating a focus on relationships or compositions

    oai_token-act-pair · gpt-4o-mini
    Triggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    ctigges/pythia-70m-deduped__res-sm_processed/2-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.2.hook_resid_post
    Hook Layer
    2
    Architecture
    standard
    Context Size
    128
    Dataset
    EleutherAI/the_pile_deduplicated
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Neuron Alignment
    Index
    Value
    % of L₁
    98
    +0.17
    0.9%
    407
    +0.11
    0.6%
    240
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    457
    +0.17
    0.07
    404
    +0.11
    0.07
    81
    +0.11
    0.07
    Negative Logits
    pson
    -1.62
    mann
    -1.43
    meter
    -1.43
    pler
    -1.42
    berg
    -1.34
    clock
    -1.34
     1200
    -1.30
     instructions
    -1.29
     minute
    -1.29
    amiento
    -1.28
    POSITIVE LOGITS
     affairs
    1.60
     whatsoever
    1.59
    ()).
    1.47
     such
    1.46
    ès
    1.44
    Australia
    1.42
     deserve
    1.39
    Australian
    1.39
    anth
    1.36
    ¿½
    1.32
    Activations Density 0.490%

    No Known Activations