© 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. 8735
    Prev
    Next
    INDEX
    Explanations

    affirmative statements or strong assertions

    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/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
    Neuron Alignment
    Index
    Value
    % of L₁
    396
    +0.15
    0.8%
    386
    +0.11
    0.6%
    306
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    386
    +0.15
    0.03
    396
    +0.11
    0.04
    11
    +0.11
    0.04
    Negative Logits
    places
    -1.73
    forms
    -1.71
    fulness
    -1.55
    ños
    -1.52
    here
    -1.47
    ème
    -1.41
    ception
    -1.40
     life
    -1.39
    place
    -1.39
    ocular
    -1.38
    POSITIVE LOGITS
    amilies
    2.46
    riends
    2.33
    ı
    2.28
    lickr
    2.27
    erguson
    2.25
    irm
    2.21
    requently
    2.17
    letcher
    2.16
    ighting
    2.15
    erdinand
    2.11
    Activations Density 0.091%

    No Known Activations