© 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. 15354
    Prev
    Next
    INDEX
    Explanations

    comparative phrases related to plans, options, and choices

    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/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
    Neuron Alignment
    Index
    Value
    % of L₁
    397
    +0.14
    0.8%
    260
    +0.13
    0.7%
    212
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    260
    +0.14
    0.09
    212
    +0.13
    0.07
    240
    +0.12
    0.00
    Negative Logits
    laughter
    -1.70
    lings
    -1.49
    lament
    -1.48
    bian
    -1.43
    alc
    -1.43
    olk
    -1.43
    rfloor
    -1.40
    ceral
    -1.38
    Const
    -1.36
    Footnote
    -1.33
    POSITIVE LOGITS
     best
    3.06
    ©
    2.74
    Ĥ
    2.74
    ĸ
    2.71
     suitable
    2.69
    §
    2.50
    Ļ
    2.48
     optimal
    2.47
    ª
    2.46
     optimum
    2.44
    Activations Density 1.529%

    No Known Activations