© 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 EvalsExportsAPI 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. 6562
    Prev
    Next
    INDEX
    Explanations

    the repetition of the word "keep" 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
    543
    +0.12
    0.6%
    566
    +0.12
    0.6%
    184
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    441
    +0.12
    0.04
    533
    +0.12
    0.04
    246
    +0.10
    0.04
    Negative Logits
    obi
    -0.95
    ode
    -0.78
    ella
    -0.77
    esome
    -0.77
    iaries
    -0.77
    anasia
    -0.74
    rs
    -0.73
    ology
    -0.73
    orer
    -0.72
    oret
    -0.71
    POSITIVE LOGITS
     tabs
    0.78
     pace
    0.74
     plug
    0.70
     faith
    0.70
     pushing
    0.67
     footing
    0.65
     practicing
    0.64
     tongues
    0.64
     resisting
    0.63
     steering
    0.63
    Activations Density 0.097%

    No Known Activations