© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 1-RES-SM
    11. 29603
    Prev
    Next
    INDEX
    Explanations

    blocks of code or statement closures in programming syntax

    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/1-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.1.hook_resid_post
    Hook Layer
    1
    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
    478
    +0.15
    0.9%
    410
    +0.13
    0.7%
    129
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    8
    +0.15
    0.02
    129
    +0.13
    0.01
    460
    +0.13
    0.02
    Negative Logits
    ense
    -1.83
    aurus
    -1.70
    clair
    -1.63
    ty
    -1.63
    ivos
    -1.56
     capacity
    -1.55
    atum
    -1.51
    ater
    -1.49
    ishers
    -1.48
     inventory
    -1.48
    POSITIVE LOGITS
     charset
    1.63
     hence
    1.61
     fact
    1.60
     here
    1.53
     consequently
    1.49
     stark
    1.46
     moreover
    1.46
    zzle
    1.45
     GOOGLE
    1.45
     therefore
    1.44
    Activations Density 0.089%

    No Known Activations