© 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. 8957
    Prev
    Next
    INDEX
    Explanations

    instances of the word "can" and its variations, indicating ability or potential

    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
    78
    +0.25
    1.4%
    238
    +0.12
    0.7%
    148
    +0.11
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    420
    +0.25
    0.07
    238
    +0.12
    0.03
    151
    +0.11
    0.04
    Negative Logits
    ĥ½
    -1.80
    ĨĴ
    -1.69
    ľ
    -1.66
    ı
    -1.54
    ¡
    -1.47
    origin
    -1.42
    enced
    -1.40
    geries
    -1.39
    les
    -1.38
    heated
    -1.35
    POSITIVE LOGITS
     spare
    1.78
     think
    1.77
     imagine
    1.65
     deduce
    1.61
     remember
    1.60
     find
    1.56
     guess
    1.54
     learn
    1.54
     behold
    1.48
     afford
    1.47
    Activations Density 0.385%

    No Known Activations