© 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. MLP Post
    9. 2-MLP-SM
    11. 27074
    Prev
    Next
    INDEX
    Explanations

    the word "can" indicating potential or capability

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    ctigges/pythia-70m-deduped__mlp-sm_processed/2-mlp-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_mlp_out
    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
    256
    +0.12
    0.7%
    432
    +0.11
    0.6%
    169
    +0.10
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    256
    +0.12
    0.06
    88
    +0.11
    0.04
    297
    +0.10
    0.05
    Negative Logits
    eren
    -1.82
    olves
    -1.57
     everything
    -1.56
     beings
    -1.53
    comes
    -1.51
     all
    -1.50
     complete
    -1.49
    esome
    -1.49
     open
    -1.48
    always
    -1.47
    POSITIVE LOGITS
    »
    1.69
    Ł
    1.62
    vä
    1.61
    Verlag
    1.59
    posium
    1.59
    forum
    1.58
     parl
    1.57
    Ķ
    1.57
    áĢ¸
    1.55
    Ĵ
    1.54
    Activations Density 0.222%

    No Known Activations