© 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. 14358
    Prev
    Next
    INDEX
    Explanations

    instances of the word "ge" in various contexts

    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
    73
    +0.11
    0.6%
    434
    +0.11
    0.6%
    219
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    78
    +0.11
    0.02
    73
    +0.11
    0.01
    386
    +0.11
    0.01
    Negative Logits
     belong
    -1.81
     belongs
    -1.59
     picture
    -1.56
     possibly
    -1.56
     ends
    -1.55
     ending
    -1.54
     end
    -1.53
     better
    -1.52
     revenge
    -1.49
    ustration
    -1.49
    POSITIVE LOGITS
    vity
    2.44
    letons
    2.11
    hog
    1.84
    stones
    1.81
    lemental
    1.75
    backed
    1.74
    à±ģ
    1.70
    orgetown
    1.69
    ños
    1.69
    remia
    1.68
    Activations Density 0.149%

    No Known Activations