© 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 49152
    9. 8-RES_FS49152-JB
    11. 26817
    Prev
    Next
    INDEX
    Explanations

    instances of the word "impossible."

    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
    620
    +0.13
    0.6%
    526
    +0.12
    0.6%
    717
    +0.12
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    620
    +0.13
    0.01
    232
    +0.12
    0.01
    202
    +0.12
    0.01
    Negative Logits
     Seym
    -0.84
    ADRA
    -0.83
    çĦ
    -0.79
    sbm
    -0.76
     Amend
    -0.74
    creen
    -0.72
     differe
    -0.71
     Hath
    -0.70
     warr
    -0.69
    ccording
    -0.69
    POSITIVE LOGITS
    ishly
    1.07
    romptu
    0.96
    ulse
    0.93
    ossible
    0.92
    ossibly
    0.91
    ressive
    0.91
    ager
    0.91
    ression
    0.90
    inson
    0.89
    agers
    0.87
    Activations Density 0.003%

    No Known Activations