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 3072
    5. 8-RES_FS3072-JB
    6. 2937
    Prev
    Next
    INDEX
    Explanations

    technical information or updates related to a particular product or service

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs3072-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    3,072
    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
    333
    +0.12
    0.5%
    756
    +0.11
    0.5%
    219
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    665
    +0.12
    0.14
    756
    +0.11
    -0.04
    333
    +0.11
    0.07
    Negative Logits
    Born
    -0.78
     hated
    -0.77
    escent
    -0.75
    utterstock
    -0.75
    famous
    -0.74
    Die
    -0.73
     badass
    -0.73
     fame
    -0.73
     gull
    -0.72
     immortal
    -0.72
    POSITIVE LOGITS
     consultations
    1.30
     briefed
    1.25
     consultation
    1.21
     feasibility
    1.18
     inquiries
    1.17
     contingency
    1.16
     pending
    1.10
     enqu
    1.10
     clarification
    1.10
     lia
    1.07
    Activations Density 3.357%

    No Known Activations