© 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
    2. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    3. Pythia-70M-Deduped
    4. Residuals
    5. 4-RES-SM
    6. 11717
    Prev
    Next
    INDEX
    Explanations

    the word "are" and its variations, focusing on the presence of the contraction "you're"

    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
    150
    +0.12
    0.7%
    426
    +0.12
    0.7%
    214
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    214
    +0.12
    0.07
    426
    +0.12
    0.07
    150
    +0.12
    0.07
    Negative Logits
    Ĭ
    -3.50
    ¥
    -3.31
    »¿
    -3.29
    º
    -3.21
    Ĥ¬
    -3.19
    Ħ
    -3.09
    ł
    -3.09
    ª
    -3.08
    Ķ
    -3.06
    ģ
    -2.97
    POSITIVE LOGITS
     interested
    2.88
     lucky
    2.52
     acquainted
    2.35
     aware
    2.33
     unsure
    2.32
     willing
    2.32
     kidding
    2.20
     fortunate
    2.19
     happy
    2.17
     afraid
    2.10
    Activations Density 0.430%

    No Known Activations