© 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 49152
    9. 8-RES_FS49152-JB
    11. 45723
    Prev
    Next
    INDEX
    Explanations

    the word "which."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    430
    +0.13
    0.6%
    165
    +0.11
    0.5%
    393
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    655
    +0.13
    0.06
    109
    +0.11
    0.05
    757
    +0.11
    0.04
    Negative Logits
     VIDEOS
    -0.78
    athi
    -0.67
    Bas
    -0.66
    MENTS
    -0.65
    BOX
    -0.64
    nor
    -0.63
    STE
    -0.61
     Waiting
    -0.60
     Editors
    -0.59
    MENT
    -0.58
    POSITIVE LOGITS
     translates
    1.10
     consists
    1.00
     originated
    0.98
     comprises
    0.96
     specializes
    0.95
     incidentally
    0.95
     represents
    0.93
     resulted
    0.92
     consisted
    0.92
     operates
    0.91
    Activations Density 0.097%

    No Known Activations