© 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. 15088
    Prev
    Next
    INDEX
    Explanations

    occurrences of underscores in the text

    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/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
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    21
    +0.27
    1.6%
    78
    +0.20
    1.2%
    156
    +0.17
    1.1%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    110
    +0.27
    0.06
    295
    +0.20
    0.04
    78
    +0.17
    0.00
    Negative Logits
    itis
    -1.58
    iated
    -1.40
    frastructure
    -1.35
     MW
    -1.33
    itian
    -1.31
     Sheet
    -1.30
    verter
    -1.29
    ket
    -1.28
     subsequent
    -1.27
    otrophic
    -1.26
    POSITIVE LOGITS
    ¡
    2.47
    ¸
    2.47
    Ĺ
    2.41
    ı
    2.41
    ¶
    2.35
    Ń
    2.35
    Ĵ
    2.25
    ·
    2.24
    ģ
    2.19
    IJ
    2.17
    Activations Density 0.260%

    No Known Activations