© 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. 45539
    Prev
    Next
    INDEX
    Explanations

    phrases indicating quantity or numerical references

    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
    241
    +0.11
    0.5%
    450
    +0.10
    0.5%
    42
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    242
    +0.11
    0.02
    409
    +0.10
    0.02
    557
    +0.10
    0.02
    Negative Logits
     Cosponsors
    -0.71
     Fenrir
    -0.68
    vir
    -0.63
    âĺ
    -0.61
    aque
    -0.60
     Tanz
    -0.59
     sensit
    -0.59
    helle
    -0.58
    mare
    -0.57
    pots
    -0.57
    POSITIVE LOGITS
    aneers
    0.67
     nine
    0.67
     eight
    0.66
     seven
    0.63
     six
    0.61
     four
    0.61
     three
    0.60
     attempted
    0.60
     consecutive
    0.59
    agements
    0.59
    Activations Density 0.039%

    No Known Activations