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 1536
    5. 8-RES_FS1536-JB
    6. 858
    Prev
    Next
    INDEX
    Explanations

    mentions of politics and corruption

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    288
    +0.15
    0.7%
    546
    +0.12
    0.6%
    679
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.15
    0.11
    546
    +0.12
    0.08
    325
    +0.11
    0.07
    Negative Logits
     Cancel
    -0.90
    actory
    -0.84
    imates
    -0.77
    wolves
    -0.74
    oa
    -0.72
    amination
    -0.71
    tered
    -0.70
     Takeru
    -0.70
    wered
    -0.70
    olen
    -0.68
    POSITIVE LOGITS
     correctness
    1.37
    eering
    0.97
     activism
    0.90
     intrig
    0.89
     pund
    0.88
     appoint
    0.88
     rhetoric
    0.87
     clout
    0.82
     affiliation
    0.81
     affili
    0.81
    Activations Density 1.441%

    No Known Activations