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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 49152
    5. 8-RES_FS49152-JB
    6. 41324
    Prev
    Next
    INDEX
    Explanations

    the word "ever" in various contexts and forms

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    98
    +0.16
    0.7%
    275
    +0.11
    0.5%
    446
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    579
    +0.16
    0.02
    98
    +0.11
    0.02
    749
    +0.10
    0.01
    Negative Logits
    OUR
    -0.66
    imilar
    -0.64
     Parables
    -0.64
    emia
    -0.64
    OSP
    -0.63
    vier
    -0.60
    toc
    -0.60
    INA
    -0.59
    Narr
    -0.59
    externalActionCode
    -0.59
    POSITIVE LOGITS
     recons
    0.87
     someday
    0.81
    theless
    0.78
     ceases
    0.78
     existed
    0.78
     doubted
    0.77
     wished
    0.76
     revisit
    0.76
    ught
    0.74
     again
    0.72
    Activations Density 0.015%

    No Known Activations