© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 1-RES-SM
    11. 25313
    Prev
    Next
    INDEX
    Explanations

    phrases related to "up" or the action of lifting/improving

    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/1-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.1.hook_resid_post
    Hook Layer
    1
    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
    117
    +0.13
    0.7%
    209
    +0.10
    0.6%
    253
    +0.10
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    12
    +0.13
    0.02
    319
    +0.10
    0.03
    317
    +0.10
    0.02
    Negative Logits
    ŀ
    -2.85
    į
    -2.69
    ·
    -2.67
    ®
    -2.63
    ¸
    -2.57
    º
    -2.53
    ½
    -2.46
    ±
    -2.41
    İ
    -2.38
    Ĵ
    -2.38
    POSITIVE LOGITS
     signing
    1.56
     recruiting
    1.55
     numerically
    1.50
     becoming
    1.49
     graduates
    1.43
     ever
    1.42
    uration
    1.42
     cellar
    1.41
    ocre
    1.39
     successfully
    1.38
    Activations Density 0.335%

    No Known Activations