© 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. 27768
    Prev
    Next
    INDEX
    Explanations

    instances of the word "understand" and its variations, indicating a focus on comprehension and insight

    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
    198
    +0.12
    0.6%
    470
    +0.11
    0.6%
    115
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    394
    +0.12
    0.03
    246
    +0.11
    0.03
    179
    +0.11
    0.03
    Negative Logits
    kins
    -1.58
    ños
    -1.55
     Ltd
    -1.55
    chel
    -1.49
    ainen
    -1.47
    ember
    -1.47
     forth
    -1.46
    eks
    -1.46
     unpublished
    -1.45
     Archive
    -1.44
    POSITIVE LOGITS
     how
    2.15
    ably
    2.15
     why
    2.11
    ®
    1.79
     language
    1.75
    Ļª
    1.67
     gaps
    1.65
    ©
    1.65
    Į
    1.57
     enough
    1.54
    Activations Density 0.176%

    No Known Activations