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 24576
    5. 8-RES_FS24576-JB
    6. 21765
    Prev
    Next
    INDEX
    Explanations

    information related to conversations and interactions between specific individuals

    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
    87
    +0.44
    2.3%
    481
    +0.14
    0.7%
    283
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    87
    +0.44
    -0.01
    665
    +0.14
    0.08
    526
    +0.08
    -0.02
    Negative Logits
    ggles
    -0.64
     wonderfully
    -0.60
     leaps
    -0.60
    ibaba
    -0.59
    uitive
    -0.58
     beautifully
    -0.55
    pires
    -0.55
    urches
    -0.55
    chieve
    -0.54
     ingenious
    -0.53
    POSITIVE LOGITS
     incrim
    0.70
     confidentiality
    0.66
     impartial
    0.64
     reconsider
    0.61
     pending
    0.61
     confidential
    0.59
     retribution
    0.58
     revoked
    0.56
     consulate
    0.55
     libel
    0.55
    Activations Density 1.300%

    No Known Activations