© 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. 14273
    Prev
    Next
    INDEX
    Explanations

    patterns in programming structures and tasks within code

    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
    270
    +0.12
    0.7%
    406
    +0.12
    0.7%
    422
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    233
    +0.12
    0.08
    96
    +0.12
    0.04
    469
    +0.11
    0.09
    Negative Logits
    ench
    -1.51
    quier
    -1.50
    obar
    -1.49
    erved
    -1.49
    idian
    -1.47
    upp
    -1.43
    ousand
    -1.42
    alf
    -1.42
    onomic
    -1.42
     Equal
    -1.41
    POSITIVE LOGITS
    ĻĤ
    2.07
    ©
    2.03
    ľĵ
    2.02
    ¼
    1.90
     lenses
    1.79
     hats
    1.73
    ¿
    1.67
    ĩ
    1.67
    ª
    1.67
     hat
    1.66
    Activations Density 3.918%

    No Known Activations