© 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. 2-RES-SM
    11. 31200
    Prev
    Next
    INDEX
    Explanations

    phrases related to awareness and understanding of various subjects or issues

    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
    176
    +0.14
    0.8%
    358
    +0.13
    0.7%
    136
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    406
    +0.14
    0.12
    358
    +0.13
    0.12
    176
    +0.13
    0.12
    Negative Logits
    ¯
    -1.86
     unfolded
    -1.79
    ģ
    -1.67
    liest
    -1.66
    Ŀ
    -1.65
    arf
    -1.59
    ħ
    -1.53
    ª
    -1.53
    ching
    -1.51
    mith
    -1.51
    POSITIVE LOGITS
    ductive
    1.65
     APPEALS
    1.60
    ably
    1.55
    claimed
    1.52
     wrongdoing
    1.50
     punishment
    1.46
     intentions
    1.43
     affairs
    1.43
    adesh
    1.43
     '_
    1.42
    Activations Density 1.247%

    No Known Activations