© 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. 41659
    Prev
    Next
    INDEX
    Explanations

    instances of the word "but", indicating a contrast or turning point in the narrative

    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
    598
    +0.10
    0.5%
    273
    +0.10
    0.5%
    501
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    598
    +0.10
    0.05
    501
    +0.10
    0.06
    157
    +0.09
    0.05
    Negative Logits
    obyl
    -0.74
    pointer
    -0.71
    kie
    -0.71
    ses
    -0.70
    ãģĨ
    -0.66
    PA
    -0.64
    zona
    -0.62
    heads
    -0.60
    amus
    -0.60
    irms
    -0.59
    POSITIVE LOGITS
     alas
    1.23
    tons
    1.13
     beware
    1.04
     unlike
    0.98
     whereas
    0.91
    chers
    0.87
     strangely
    0.79
     unfortunately
    0.78
     despite
    0.76
     fortunately
    0.75
    Activations Density 0.103%

    No Known Activations