© 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. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 17778
    Prev
    Next
    INDEX
    Explanations

    references to recent events or developments

    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
    744
    +0.11
    0.5%
    592
    +0.11
    0.5%
    306
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    306
    +0.11
    0.04
    115
    +0.11
    0.03
    744
    +0.10
    0.03
    Negative Logits
     idols
    -0.85
     brakes
    -0.80
    seekers
    -0.76
    amines
    -0.75
     dinosaurs
    -0.74
    ammers
    -0.74
     apes
    -0.73
    aters
    -0.72
     rails
    -0.71
     detectors
    -0.70
    POSITIVE LOGITS
     statement
    0.99
     interview
    0.95
    statement
    0.93
     article
    0.90
     survey
    0.88
     conversation
    0.88
     excerpt
    0.87
    lished
    0.85
     affidavit
    0.85
     report
    0.83
    Activations Density 0.080%

    No Known Activations