© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 8630
    Prev
    Next
    INDEX
    Explanations

    phrases related to breaking or disruption

    oai_token-act-pair · gpt-4o-mini
    Triggered 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
    Neuron Alignment
    Index
    Value
    % of L₁
    337
    +0.11
    0.6%
    194
    +0.11
    0.6%
    62
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    62
    +0.11
    0.02
    198
    +0.11
    0.04
    194
    +0.11
    0.02
    Negative Logits
    ¦
    -1.84
     CURIAM
    -1.63
    chio
    -1.53
    iente
    -1.51
    :`
    -1.51
    ientes
    -1.50
    allas
    -1.47
     name
    -1.46
     BASIS
    -1.43
    izumab
    -1.42
    POSITIVE LOGITS
    neck
    2.31
     apart
    2.18
    waters
    2.07
    fast
    2.02
    board
    2.02
     loose
    2.00
    down
    1.97
    hearted
    1.96
    through
    1.94
    out
    1.91
    Activations Density 0.316%

    No Known Activations