© 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. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 2552
    Prev
    Next
    INDEX
    Explanations

    negative or controversial information about celebrities

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    373
    +0.13
    0.6%
    314
    +0.10
    0.5%
    436
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    290
    +0.13
    0.02
    373
    +0.10
    -0.01
    432
    +0.10
    0.02
    Negative Logits
     nailed
    -0.65
    ashtra
    -0.64
    utics
    -0.60
    estic
    -0.60
     Flavoring
    -0.57
    ELF
    -0.57
    eeks
    -0.56
    icans
    -0.56
    ateurs
    -0.55
    utic
    -0.55
    POSITIVE LOGITS
    docker
    0.74
    ikuman
    0.73
    BuyableInstoreAndOnline
    0.73
    bum
    0.71
    Ô
    0.69
    Downloadha
    0.68
    oscope
    0.68
    opal
    0.65
    andi
    0.65
    ulum
    0.62
    Activations Density 0.058%

    No Known Activations