© 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. 1-RES-SM
    11. 24037
    Prev
    Next
    INDEX
    Explanations

    the word "was" in various contexts

    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/1-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.1.hook_resid_post
    Hook Layer
    1
    Architecture
    standard
    Context Size
    128
    Dataset
    EleutherAI/the_pile_deduplicated
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Neuron Alignment
    Index
    Value
    % of L₁
    23
    +0.42
    2.6%
    32
    +0.15
    0.9%
    410
    +0.14
    0.8%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    348
    +0.42
    0.11
    32
    +0.15
    0.10
    23
    +0.14
    0.10
    Negative Logits
    anno
    -2.04
    pieces
    -1.95
    works
    -1.93
    facts
    -1.92
    #.
    -1.89
    endas
    -1.82
    scope
    -1.82
    params
    -1.81
    #:
    -1.80
    msgid
    -1.80
    POSITIVE LOGITS
    hes
    2.13
    hers
    2.10
     unsuccessful
    1.74
     discontinued
    1.72
     supposed
    1.72
      ↵
    1.61
     gradual
    1.58
     always
    1.57
     successful
    1.56
     respectively
    1.54
    Activations Density 0.421%

    No Known Activations