© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI 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. 11766
    Prev
    Next
    INDEX
    Explanations

    phrases that convey comparisons or relative evaluations

    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
    240
    +0.14
    0.8%
    21
    +0.13
    0.7%
    375
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    375
    +0.14
    0.03
    402
    +0.13
    0.04
    102
    +0.13
    0.03
    Negative Logits
    ĥ½
    -1.85
    ed
    -1.78
    Ķ
    -1.71
    edly
    -1.54
    Ŀ
    -1.51
    IJ
    -1.50
    Ĩ
    -1.50
    Į
    -1.50
    ¥
    -1.49
    ī
    -1.49
    POSITIVE LOGITS
     inexpensive
    2.23
     harmless
    2.16
     easy
    2.11
     few
    1.92
     stable
    1.85
     tame
    1.81
     uncommon
    1.79
     straightforward
    1.78
     unchanged
    1.76
     rare
    1.72
    Activations Density 0.108%

    No Known Activations