© 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. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    5. GPT2-Small
    7. Residual Post - 32k
    9. 8-RES_POST_32K-OAI
    11. 4180
    Prev
    Next
    INDEX
    Explanations

    occurrences of the word "that" in various contexts

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.8.hook_resid_post
    Hook Layer
    8
    Architecture
    standard
    Context Size
    64
    Dataset
    Skylion007/openwebtext
    Activation Function
    topk
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    459
    +0.14
    0.6%
    47
    +0.09
    0.4%
    682
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    172
    +0.14
    0.03
    393
    +0.09
    0.04
    203
    +0.08
    0.03
    Negative Logits
    maybe
    -0.72
    MQ
    -0.66
    walk
    -0.65
    Plus
    -0.64
    sort
    -0.64
    ulton
    -0.63
     Mori
    -0.62
    ission
    -0.61
     suggestion
    -0.61
    lins
    -0.61
    POSITIVE LOGITS
     stays
    0.77
    egg
    0.68
     adequate
    0.66
     remains
    0.65
     unborn
    0.65
     passes
    0.62
    tains
    0.61
     lid
    0.61
     loopholes
    0.61
     proper
    0.60
    Activations Density 0.200%

    No Known Activations