© 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. 2-RES-SM
    11. 14097
    Prev
    Next
    INDEX
    Explanations

    references to data encoding formats

    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
    283
    +0.13
    0.7%
    197
    +0.11
    0.6%
    271
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    283
    +0.13
    0.01
    180
    +0.11
    0.01
    370
    +0.11
    0.01
    Negative Logits
     killing
    -1.95
    !’
    -1.57
    oned
    -1.57
    ’.
    -1.55
    duct
    -1.55
     chasing
    -1.54
     successful
    -1.53
     winning
    -1.52
     rescue
    -1.51
     beating
    -1.51
    POSITIVE LOGITS
     languages
    2.19
     headers
    1.81
    ŀ
    1.68
    anguage
    1.66
    °
    1.63
     Language
    1.58
     BGCOLOR
    1.55
     UTF
    1.51
    imeters
    1.50
    cloth
    1.48
    Activations Density 0.030%

    No Known Activations