© 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. 26122
    Prev
    Next
    INDEX
    Explanations

    instances of the term "Col" followed by numbers, indicating specific classifications or categories

    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
    156
    +0.17
    1.0%
    164
    +0.16
    0.9%
    350
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    164
    +0.17
    0.03
    350
    +0.16
    0.03
    410
    +0.12
    0.03
    Negative Logits
     cause
    -1.61
     interests
    -1.52
    alus
    -1.47
     longer
    -1.47
     sure
    -1.43
     seizure
    -1.42
    ĵ
    -1.39
    rapeut
    -1.39
     speedy
    -1.38
    chell
    -1.38
    POSITIVE LOGITS
    acted
    1.76
    iff
    1.58
    ometric
    1.53
    neck
    1.53
    stick
    1.52
    leen
    1.50
    nr
    1.48
    stone
    1.47
    isco
    1.46
    lected
    1.46
    Activations Density 0.100%

    No Known Activations