© 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. 2-RES-SM
    11. 32011
    Prev
    Next
    INDEX
    Explanations

    dates referenced in the text

    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
    43
    +0.20
    1.1%
    477
    +0.14
    0.8%
    396
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    43
    +0.20
    0.03
    477
    +0.14
    0.02
    82
    +0.13
    0.02
    Negative Logits
    ¢
    -2.16
    Ĥ¬
    -2.08
    ĨĴ
    -2.02
    ĸ´
    -1.82
    ¶
    -1.82
    Ŀ
    -1.82
    ĩ
    -1.79
    Ĺ
    -1.77
    ¯
    -1.74
    ½
    -1.73
    POSITIVE LOGITS
    berry
    1.97
    ford
    1.65
    amer
    1.55
    alan
    1.54
     deadline
    1.54
    shaw
    1.50
    gins
    1.49
    berries
    1.49
    inian
    1.49
    bell
    1.48
    Activations Density 0.033%

    No Known Activations