© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 2-RES-SM
    11. 12106
    Prev
    Next
    INDEX
    Explanations

    references to colors, particularly shades of white

    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/2-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.2.hook_resid_post
    Hook Layer
    2
    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
    156
    +0.20
    1.1%
    466
    +0.16
    0.9%
    250
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    250
    +0.20
    0.03
    446
    +0.16
    0.02
    466
    +0.13
    0.03
    Negative Logits
    gio
    -1.58
    onal
    -1.56
    arily
    -1.54
     planning
    -1.50
    hematic
    -1.50
    rid
    -1.48
    ANC
    -1.48
    prud
    -1.47
    lying
    -1.46
    aph
    -1.46
    POSITIVE LOGITS
    ford
    1.89
    hurst
    1.88
    matter
    1.82
    washed
    1.77
    smith
    1.66
    cloth
    1.65
    buntu
    1.62
     Matter
    1.62
    fish
    1.61
    shirts
    1.60
    Activations Density 0.102%

    No Known Activations