© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI 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. 28827
    Prev
    Next
    INDEX
    Explanations

    programming language identifiers and code structure 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/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
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    156
    +0.20
    1.2%
    23
    +0.16
    0.9%
    263
    +0.15
    0.9%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    156
    +0.20
    0.68
    111
    +0.16
    0.69
    71
    +0.15
    0.63
    Negative Logits
    ème
    -1.75
    "}](#
    -1.68
     thesis
    -1.53
    journal
    -1.46
     argument
    -1.46
    STA
    -1.43
    CIT
    -1.41
    ле
    -1.40
    ikipedia
    -1.38
    VERTISEMENT
    -1.36
    POSITIVE LOGITS
    keit
    1.70
     nights
    1.60
     respectively
    1.52
     outright
    1.50
    lord
    1.36
    thing
    1.34
    ously
    1.32
     privileges
    1.30
    filled
    1.28
    uh
    1.28
    Activations Density 5.423%

    No Known Activations