© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    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. 9573
    Prev
    Next
    INDEX
    Explanations

    descriptive passages containing complex or detailed explanations

    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
    447
    +0.14
    0.6%
    628
    +0.10
    0.4%
    425
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    58
    +0.14
    0.03
    190
    +0.10
    0.03
    647
    +0.09
    0.03
    Negative Logits
    rawdownloadcloneembedreportprint
    -0.90
     dare
    -0.68
    proc
    -0.66
    .?
    -0.64
    odore
    -0.63
    kay
    -0.62
     sqor
    -0.62
    Quote
    -0.61
     Quote
    -0.60
    Ĥª
    -0.60
    POSITIVE LOGITS
     foreground
    0.66
    )=(
    0.62
    âĨĴ
    0.61
     comprising
    0.59
     vs
    0.58
     consisting
    0.58
     consists
    0.57
     consisted
    0.57
     âĨĴ
    0.55
     Clifford
    0.54
    Activations Density 0.353%

    No Known Activations