© 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. 4323
    Prev
    Next
    INDEX
    Explanations

    descriptions related to personal experiences and emotions

    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
    362
    +0.18
    0.9%
    393
    +0.17
    0.8%
    481
    +0.13
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    362
    +0.18
    0.15
    393
    +0.17
    0.11
    608
    +0.13
    0.12
    Negative Logits
     thereby
    -0.91
    LEASE
    -0.83
     aimed
    -0.80
    successfully
    -0.78
    exempt
    -0.78
     according
    -0.76
    iffe
    -0.75
     namely
    -0.75
    ashington
    -0.75
    verage
    -0.74
    POSITIVE LOGITS
     slightest
    1.34
     coolest
    1.25
     smell
    1.11
     thrill
    1.11
     simplest
    1.10
     smallest
    1.08
     occasional
    1.07
     smartest
    1.06
     whole
    1.05
     dreaded
    1.05
    Activations Density 0.689%

    No Known Activations