© 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. 4-RES-SM
    11. 14000
    Prev
    Next
    INDEX
    Explanations

    closing braces and specific formatting elements in structured text inputs

    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/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
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    307
    +0.18
    1.1%
    306
    +0.17
    1.1%
    78
    +0.17
    1.1%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    157
    +0.18
    0.01
    348
    +0.17
    0.01
    34
    +0.17
    0.01
    Negative Logits
     backgrounds
    -1.45
    going
    -1.37
    hold
    -1.36
    dom
    -1.34
     insulator
    -1.34
    rive
    -1.33
    ine
    -1.30
    craft
    -1.29
     contractor
    -1.28
     background
    -1.25
    POSITIVE LOGITS
    ĥ½
    4.25
    IJ
    4.23
    ł
    4.07
    ¡
    4.03
    Ĭ
    3.97
    ij
    3.94
    ·
    3.93
    ĻĤ
    3.90
    ½
    3.89
    ¤
    3.76
    Activations Density 0.020%

    No Known Activations