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

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 2023
    Prev
    Next
    INDEX
    Explanations

    comparisons using the word "than"

    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
    655
    +0.16
    0.8%
    432
    +0.10
    0.5%
    724
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    655
    +0.16
    0.07
    724
    +0.10
    0.05
    432
    +0.10
    0.04
    Negative Logits
    ilic
    -0.68
    Creat
    -0.67
     POV
    -0.63
    RAW
    -0.62
    bies
    -0.61
     seriousness
    -0.61
    actions
    -0.60
     Directions
    -0.60
    oft
    -0.60
    OLOG
    -0.58
    POSITIVE LOGITS
     tripled
    1.21
     doubled
    1.11
     doubling
    1.05
     twice
    0.98
     quadru
    0.88
     half
    0.82
     600
    0.81
    whelming
    0.81
     400
    0.80
     550
    0.80
    Activations Density 0.069%

    No Known Activations