© 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. 8670
    Prev
    Next
    INDEX
    Explanations

    numerical data related to measurements or statistics

    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
    437
    +0.10
    0.5%
    169
    +0.10
    0.5%
    586
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    418
    +0.10
    0.04
    409
    +0.10
    0.04
    165
    +0.10
    0.03
    Negative Logits
     bye
    -0.67
    reads
    -0.65
    chin
    -0.64
     Ruk
    -0.62
    hani
    -0.60
    team
    -0.57
    boarding
    -0.57
     sockets
    -0.57
     brig
    -0.56
    leneck
    -0.55
    POSITIVE LOGITS
    000
    1.00
    merce
    0.79
    ancies
    0.72
     Mens
    0.72
    Ń·
    0.70
    istas
    0.69
    etheless
    0.66
    utsche
    0.66
     laun
    0.66
    tions
    0.65
    Activations Density 0.164%

    No Known Activations