© 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 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. 10303
    Prev
    Next
    INDEX
    Explanations

    formal language typically found in legal documents

    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
    159
    +0.14
    0.8%
    157
    +0.14
    0.8%
    320
    +0.13
    0.8%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    157
    +0.14
    0.04
    222
    +0.14
    0.03
    359
    +0.13
    0.03
    Negative Logits
     respectively
    -1.95
    unnumbered
    -1.76
    ))$.
    -1.75
     etc
    -1.52
    soever
    -1.52
    ))$
    -1.47
    ))$,
    -1.45
    yan
    -1.36
    vae
    -1.36
     counterparts
    -1.35
    POSITIVE LOGITS
     myself
    1.69
    hift
    1.47
     upstream
    1.45
     emergency
    1.42
     ourselves
    1.38
     applicable
    1.37
     office
    1.34
     appointed
    1.34
     underlying
    1.34
     intended
    1.32
    Activations Density 0.100%

    No Known Activations