© 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. 5168
    Prev
    Next
    INDEX
    Explanations

    sports-related terms such as match outcomes and gameplay tactics

    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
    481
    +0.21
    1.0%
    314
    +0.15
    0.7%
    540
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    540
    +0.21
    0.05
    531
    +0.15
    0.04
    274
    +0.12
    0.03
    Negative Logits
     incorpor
    -0.87
     trave
    -0.78
     notor
    -0.70
     millenn
    -0.66
     concess
    -0.65
     thous
    -0.64
     certific
    -0.62
     unden
    -0.62
     exha
    -0.60
     constitu
    -0.60
    POSITIVE LOGITS
    nings
    1.29
    able
    1.23
    ability
    1.22
    ings
    1.10
    cation
    1.08
    athon
    1.08
    ables
    1.07
    away
    1.06
    aways
    1.03
    over
    1.03
    Activations Density 0.207%

    No Known Activations