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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsBlogSlackPrivacy & 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. 344
    Prev
    Next
    INDEX
    Explanations

    verbs related to research, exploration, feelings, and financial support

    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
    363
    +0.11
    0.5%
    463
    +0.11
    0.5%
    231
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    150
    +0.11
    0.09
    206
    +0.11
    0.09
    382
    +0.10
    0.10
    Negative Logits
    rid
    -0.69
    ollah
    -0.67
    OG
    -0.61
     Manufacturer
    -0.61
     permitting
    -0.59
    interstitial
    -0.59
     drying
    -0.59
     Results
    -0.59
     culminating
    -0.59
    parser
    -0.57
    POSITIVE LOGITS
     whom
    1.14
    who
    0.89
    pired
    0.86
    doms
    0.78
    chool
    0.77
     benefited
    0.77
    whose
    0.76
    abouts
    0.76
    pires
    0.75
     died
    0.74
    Activations Density 6.248%

    No Known Activations