© 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. 4181
    Prev
    Next
    INDEX
    Explanations

    expressions of disapproval and intention related to family or personal relationships

    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
    373
    +0.10
    0.5%
    4
    +0.10
    0.5%
    682
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    362
    +0.10
    0.11
    616
    +0.10
    0.07
    4
    +0.10
    0.07
    Negative Logits
    artifacts
    -0.86
    elight
    -0.72
    rients
    -0.70
    idates
    -0.68
    tics
    -0.67
    gat
    -0.66
    pole
    -0.65
    itives
    -0.63
    ounters
    -0.63
    mite
    -0.62
    POSITIVE LOGITS
     permission
    1.26
     authorization
    1.02
     supervision
    1.01
     consent
    1.01
     behest
    0.99
     encouragement
    0.97
     approval
    0.96
     blessing
    0.93
     assistance
    0.92
     complicity
    0.89
    Activations Density 1.554%

    No Known Activations