© 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. 26384
    Prev
    Next
    INDEX
    Explanations

    conjunctions, particularly the word "and"

    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
    478
    +0.20
    1.1%
    153
    +0.12
    0.7%
    219
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    219
    +0.20
    0.09
    256
    +0.12
    0.08
    321
    +0.12
    0.06
    Negative Logits
    azine
    -1.58
    iol
    -1.54
    ees
    -1.54
     himself
    -1.53
    ures
    -1.50
    ains
    -1.49
    iors
    -1.45
    fire
    -1.44
    !(
    -1.42
     {#
    -1.40
    POSITIVE LOGITS
     consequent
    1.81
    alus
    1.71
     hence
    1.70
     yes
    1.67
     frankly
    1.59
     subsequent
    1.57
     rebuilding
    1.56
     consequently
    1.52
     hopefully
    1.48
     vice
    1.48
    Activations Density 0.409%

    No Known Activations