© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 14737
    Prev
    Next
    INDEX
    Explanations

    phrases related to the importance and impact of various actions or decisions

    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
    +0.11
    0.5%
    608
    +0.10
    0.5%
    150
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    608
    +0.11
    0.03
    49
    +0.10
    0.03
    665
    +0.10
    0.03
    Negative Logits
    ãĤ¦ãĤ¹
    -0.85
    door
    -0.74
    ELD
    -0.72
    netflix
    -0.72
    soever
    -0.70
    MpServer
    -0.69
    ById
    -0.69
    furt
    -0.68
    ihad
    -0.66
    holes
    -0.65
    POSITIVE LOGITS
     emphasize
    1.17
     clarify
    1.17
     emphas
    1.15
     ensure
    1.11
     dispel
    1.07
     understand
    1.05
     beware
    1.04
     note
    1.03
     remember
    1.03
     avoid
    1.02
    Activations Density 0.072%

    No Known Activations