© 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. 13762
    Prev
    Next
    INDEX
    Explanations

    instances of the word "start" in various forms

    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
    211
    +0.14
    0.8%
    78
    +0.13
    0.7%
    194
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    211
    +0.14
    0.05
    99
    +0.13
    0.03
    194
    +0.13
    0.03
    Negative Logits
    ous
    -1.89
    aries
    -1.76
    ORY
    -1.53
    matic
    -1.50
    ITY
    -1.49
    endants
    -1.49
    avier
    -1.48
    clude
    -1.44
    rous
    -1.41
    aille
    -1.40
    POSITIVE LOGITS
     date
    1.81
    tls
    1.76
    up
    1.75
    lessly
    1.75
    odontic
    1.74
     up
    1.73
     pitching
    1.71
     menu
    1.64
    ups
    1.63
    feet
    1.63
    Activations Density 0.191%

    No Known Activations