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

    Neuronpedia

    Jacobian LensNEW
    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. 2-RES-SM
    11. 9999
    Prev
    Next
    INDEX
    Explanations

    references to licenses or legal terms related to software

    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
    320
    +0.14
    0.8%
    240
    +0.12
    0.7%
    281
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    214
    +0.14
    0.01
    425
    +0.12
    0.02
    177
    +0.11
    0.01
    Negative Logits
    Ł
    -1.69
    ect
    -1.61
    į
    -1.56
    ened
    -1.56
    cyt
    -1.50
    mente
    -1.37
    vinyl
    -1.37
    stery
    -1.35
    olith
    -1.35
    bris
    -1.34
    POSITIVE LOGITS
     Redistributions
    1.68
     Exercise
    1.67
     exercise
    1.50
     States
    1.47
    andum
    1.43
     Executive
    1.43
     General
    1.41
    heed
    1.40
     §
    1.40
     permission
    1.39
    Activations Density 0.070%

    No Known Activations