© 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. 9627
    Prev
    Next
    INDEX
    Explanations

    occurrences of the letter 'w' in various contexts

    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
    279
    +0.15
    0.9%
    155
    +0.13
    0.7%
    317
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    279
    +0.15
    0.04
    290
    +0.13
    0.03
    396
    +0.12
    0.03
    Negative Logits
    unnumbered
    -1.78
    crow
    -1.59
    ULAR
    -1.59
    urally
    -1.57
    ños
    -1.54
     routes
    -1.52
     lifetime
    -1.47
    eer
    -1.47
     least
    -1.42
    gaard
    -1.41
    POSITIVE LOGITS
    elcome
    2.61
    avelength
    2.35
    arrant
    2.22
    olves
    2.19
    Ĺ
    2.19
    aiver
    2.17
    arming
    2.15
    rapped
    2.01
    avy
    1.99
    inkle
    1.98
    Activations Density 0.090%

    No Known Activations