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. 559
    Prev
    Next
    INDEX
    Explanations

    positions or roles people hold within communities

    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
    420
    +0.12
    0.6%
    579
    +0.12
    0.5%
    522
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    36
    +0.12
    0.18
    579
    +0.12
    0.14
    522
    +0.10
    0.12
    Negative Logits
    mares
    -0.83
    uden
    -0.80
    cies
    -0.80
    events
    -0.74
    ancies
    -0.73
    ispers
    -0.71
    onies
    -0.71
     patches
    -0.71
     queues
    -0.71
    encies
    -0.70
    POSITIVE LOGITS
     conduit
    1.16
     beneficiary
    1.10
     leader
    1.09
     proponent
    1.07
     negotiator
    1.06
     supporter
    1.05
     protector
    1.05
     recipient
    1.04
     performer
    1.04
     participant
    1.02
    Activations Density 4.350%

    No Known Activations