© 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 1536
    9. 8-RES_FS1536-JB
    11. 1018
    Prev
    Next
    INDEX
    Explanations

    instances where a preferred action or choice is made over an alternative

    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
    756
    +0.10
    0.4%
    414
    +0.10
    0.4%
    544
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    8
    +0.10
    0.12
    756
    +0.10
    -0.04
    49
    +0.10
    0.11
    Negative Logits
     landmark
    -0.69
     Previously
    -0.64
     milestone
    -0.61
     particularly
    -0.60
    estamp
    -0.60
    anta
    -0.60
     anniversary
    -0.59
     exceeds
    -0.58
    mentioned
    -0.57
     Earlier
    -0.57
    POSITIVE LOGITS
     merely
    1.19
     concentrate
    1.08
     simply
    1.00
    Instead
    0.90
     purely
    0.89
     relying
    0.86
     foc
    0.86
     instead
    0.85
     bland
    0.81
     focus
    0.81
    Activations Density 3.978%

    No Known Activations