© 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. 27348
    Prev
    Next
    INDEX
    Explanations

    legal terms and charges related to criminal offenses

    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
    33
    +0.11
    0.5%
    373
    +0.10
    0.5%
    727
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    458
    +0.11
    0.06
    357
    +0.10
    0.05
    516
    +0.10
    0.05
    Negative Logits
     reconc
    -0.94
     revisions
    -0.80
     refunds
    -0.78
     manuscripts
    -0.77
     tweaks
    -0.73
     portfolios
    -0.73
     printers
    -0.73
    Versions
    -0.73
     notebooks
    -0.73
     Journals
    -0.72
    POSITIVE LOGITS
     reck
    0.95
     willfully
    0.91
     committing
    0.91
    angering
    0.87
     felon
    0.86
     bodily
    0.85
     reckless
    0.84
     deadly
    0.82
     inflicted
    0.81
     knowingly
    0.81
    Activations Density 1.167%

    No Known Activations