© 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. 15128
    Prev
    Next
    INDEX
    Explanations

    adverbs and adjectives emphasizing strong qualities or conditions

    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
    202
    +0.13
    0.7%
    375
    +0.12
    0.6%
    78
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    375
    +0.13
    0.09
    202
    +0.12
    0.03
    2
    +0.11
    0.08
    Negative Logits
    Ĩ
    -2.78
    Ķ
    -2.61
    ĥ½
    -2.56
    IJ
    -2.42
    £
    -2.41
    Īĺ
    -2.34
    Ī
    -2.33
    ¡
    -2.32
    ¿½
    -2.25
    Ħ
    -2.23
    POSITIVE LOGITS
     strong
    2.48
     enjoyable
    2.48
     talented
    2.47
     loud
    2.44
     powerful
    2.40
     entertaining
    2.36
     beautiful
    2.34
     painful
    2.33
     awesome
    2.30
     easy
    2.30
    Activations Density 0.776%

    No Known Activations