© 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. 7221
    Prev
    Next
    INDEX
    Explanations

    instances of the word "simply" and its variations, suggesting a focus on straightforward or uncomplicated expressions

    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
    99
    +0.10
    0.5%
    13
    +0.10
    0.5%
    613
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    106
    +0.10
    0.03
    297
    +0.10
    0.03
    604
    +0.09
    0.03
    Negative Logits
    afort
    -0.77
     enthus
    -0.74
    amia
    -0.68
     extensively
    -0.65
     heavily
    -0.65
     seriously
    -0.65
    roach
    -0.65
    egal
    -0.61
    MH
    -0.61
     specifically
    -0.59
    POSITIVE LOGITS
     stated
    0.75
    IFIED
    0.69
     redistributed
    0.68
     shrugged
    0.67
     untrue
    0.67
    cos
    0.67
    ãģį
    0.65
     cannot
    0.65
    %]
    0.64
    speaking
    0.63
    Activations Density 0.152%

    No Known Activations