© 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. 4-RES-SM
    11. 8908
    Prev
    Next
    INDEX
    Explanations

    references to programming structures or code elements

    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/4-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.4.hook_resid_post
    Hook Layer
    4
    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
    78
    +0.19
    1.1%
    451
    +0.12
    0.7%
    339
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    461
    +0.19
    0.01
    54
    +0.12
    0.01
    78
    +0.12
    -0.00
    Negative Logits
    ership
    -1.61
     operator
    -1.46
     course
    -1.41
    eras
    -1.30
    aries
    -1.28
     guide
    -1.27
    iginally
    -1.27
     partner
    -1.26
     negatively
    -1.25
     operators
    -1.25
    POSITIVE LOGITS
    ķ
    3.85
    ª
    3.82
    ³
    3.79
    ģ
    3.79
    ij
    3.75
    ¡
    3.73
    ĨĴ
    3.73
    Ķ
    3.71
    £
    3.67
    <|outofrange|>
    3.67
    Activations Density 0.024%

    No Known Activations