Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 768
    5. 8-RES_FS768-JB
    6. 129
    Prev
    Next
    INDEX
    Explanations

    texts related to career resignations and police investigations

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    373
    +0.14
    0.6%
    493
    +0.11
    0.5%
    500
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    373
    +0.14
    -0.17
    640
    +0.11
    0.22
    470
    +0.11
    0.07
    Negative Logits
    colored
    -1.00
     Navajo
    -0.98
     favors
    -0.96
     gotten
    -0.96
     nonprofits
    -0.96
    imize
    -0.94
     DOE
    -0.92
     DOD
    -0.92
     Whitman
    -0.91
     favorably
    -0.90
    POSITIVE LOGITS
     £
    1.59
     realise
    1.51
     organising
    1.49
     centre
    1.48
     footballer
    1.48
     enqu
    1.47
     recognise
    1.47
     Whilst
    1.44
    colour
    1.42
     favourites
    1.42
    Activations Density 9.044%

    No Known Activations