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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & 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. 1529
    Prev
    Next
    INDEX
    Explanations

    technical terms related to database querying and optimization

    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
    260
    +0.10
    0.5%
    340
    +0.10
    0.5%
    45
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    45
    +0.10
    0.04
    718
    +0.10
    0.10
    283
    +0.10
    0.01
    Negative Logits
    ËĪ
    -0.72
    inventoryQuantity
    -0.66
    CHAPTER
    -0.65
     )]
    -0.63
    Optional
    -0.62
     Reserve
    -0.60
     Psychiatric
    -0.58
     Sacrifice
    -0.58
    annis
    -0.58
     Kes
    -0.58
    POSITIVE LOGITS
     results
    1.07
    RESULTS
    1.07
    success
    1.03
    result
    0.96
    results
    0.95
    successfully
    0.94
     success
    0.89
     invariably
    0.88
    failed
    0.88
     yielded
    0.87
    Activations Density 3.919%

    No Known Activations