© 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. 26086
    Prev
    Next
    INDEX
    Explanations

    articles in the text

    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
    +0.25
    1.5%
    436
    +0.20
    1.2%
    478
    +0.15
    0.9%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    436
    +0.25
    0.06
    230
    +0.20
    0.03
    321
    +0.15
    0.04
    Negative Logits
    ior
    -1.59
    idium
    -1.58
    ions
    -1.55
    enance
    -1.52
    ]{.
    -1.52
    æĶ¿
    -1.52
     Ages
    -1.50
    endor
    -1.50
    lament
    -1.43
    INESS
    -1.42
    POSITIVE LOGITS
     dozen
    1.98
     lot
    1.95
     rather
    1.94
     nice
    1.92
     handful
    1.89
     few
    1.87
    ktiv
    1.87
     doubt
    1.83
     pretty
    1.82
    Ĵ
    1.80
    Activations Density 0.220%

    No Known Activations