© 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. 7187
    Prev
    Next
    INDEX
    Explanations

    phrases or structures that create suspense or tension

    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
    720
    +0.12
    0.6%
    43
    +0.12
    0.6%
    393
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    43
    +0.12
    0.23
    720
    +0.12
    0.27
    546
    +0.12
    0.23
    Negative Logits
    izen
    -0.72
    ¬¼
    -0.70
    ocl
    -0.69
    irie
    -0.65
    chell
    -0.62
    abouts
    -0.58
    è£ħ
    -0.58
    =#
    -0.57
    sein
    -0.57
    reated
    -0.57
    POSITIVE LOGITS
     whereas
    1.43
     although
    1.41
     namely
    1.37
     albeit
    1.34
     but
    1.32
     which
    1.27
     though
    1.16
     whereby
    1.15
     thereby
    1.13
    which
    1.12
    Activations Density 2.701%

    No Known Activations