© 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 49152
    9. 8-RES_FS49152-JB
    11. 41702
    Prev
    Next
    INDEX
    Explanations

    references to flipping or switching actions and concepts

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    50
    +0.13
    0.6%
    636
    +0.12
    0.5%
    140
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    636
    +0.13
    0.03
    526
    +0.12
    0.01
    362
    +0.10
    0.02
    Negative Logits
    CLA
    -0.69
    shaw
    -0.67
    lain
    -0.67
    CLUD
    -0.66
    cliffe
    -0.64
     Heights
    -0.63
    DEF
    -0.62
    CVE
    -0.62
    interrupted
    -0.62
    NAME
    -0.61
    POSITIVE LOGITS
     coin
    0.94
     punches
    0.80
     switch
    0.79
     scales
    0.72
     torch
    0.72
     curtain
    0.72
     LED
    0.72
     LEDs
    0.72
     tide
    0.71
     dice
    0.70
    Activations Density 0.062%

    No Known Activations