© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 768
    9. 8-RES_FS768-JB
    11. 755
    Prev
    Next
    INDEX
    Explanations

    phrases related to time and history

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    636
    +0.12
    0.5%
    505
    +0.10
    0.5%
    269
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    636
    +0.12
    0.12
    505
    +0.10
    0.10
    95
    +0.10
    0.07
    Negative Logits
    urated
    -0.65
    elected
    -0.62
    overed
    -0.60
     submitting
    -0.59
    fed
    -0.58
    eming
    -0.58
    gem
    -0.58
    being
    -0.57
     Return
    -0.57
     possessing
    -0.56
    POSITIVE LOGITS
     differently
    0.97
    thin
    0.76
     rounds
    0.75
    itud
    0.74
    orously
    0.73
     havoc
    0.72
    ationally
    0.72
    ibly
    0.70
     responsibly
    0.70
     groundwork
    0.70
    Activations Density 7.366%

    No Known Activations