© 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. 10-RES_POST_32K-OAI
    11. 27156
    Prev
    Next
    INDEX
    Explanations

    references to numerical data and statistics

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.10.hook_resid_post
    Hook Layer
    10
    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
    415
    +0.12
    0.5%
    7
    +0.11
    0.5%
    756
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    488
    +0.12
    0.02
    316
    +0.11
    0.02
    373
    +0.11
    -0.02
    Negative Logits
     spectator
    -0.67
     Dragons
    -0.67
    kered
    -0.63
    âĹ¼
    -0.59
     unlaw
    -0.58
     Syri
    -0.58
    Redditor
    -0.56
    elist
    -0.55
     Georgian
    -0.54
    inka
    -0.54
    POSITIVE LOGITS
    dirty
    0.66
     CLR
    0.64
    ific
    0.62
    Ħ¢
    0.62
    little
    0.61
    20439
    0.61
    wall
    0.60
    inently
    0.59
     ¶
    0.58
     Ct
    0.57
    Activations Density 0.092%

    No Known Activations