© 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. 29847
    Prev
    Next
    INDEX
    Explanations

    occurrences of the word "debate" in various contexts

    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/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
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    58
    +0.14
    0.8%
    348
    +0.13
    0.7%
    147
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    348
    +0.14
    0.02
    58
    +0.13
    0.02
    59
    +0.12
    0.01
    Negative Logits
     direction
    -2.02
     directions
    -1.89
     whole
    -1.79
     prospective
    -1.62
    ward
    -1.60
     matter
    -1.59
     glance
    -1.58
    wards
    -1.56
    fra
    -1.56
     favor
    -1.51
    POSITIVE LOGITS
    ĻĤ
    2.37
    ¹
    2.10
    bing
    1.90
    ²
    1.86
    atable
    1.84
    isher
    1.77
    ¼
    1.74
    acle
    1.74
    »
    1.70
    ¬
    1.67
    Activations Density 0.093%

    No Known Activations