© 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. 41686
    Prev
    Next
    INDEX
    Explanations

    instances of the word "though," indicating a focus on contrast or concession within the text

    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
    373
    +0.17
    0.8%
    460
    +0.10
    0.5%
    132
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    460
    +0.17
    0.03
    8
    +0.10
    0.03
    135
    +0.10
    0.03
    Negative Logits
    anwhile
    -0.69
    ufact
    -0.66
    CLOSE
    -0.60
    idia
    -0.60
    mic
    -0.58
     Flavoring
    -0.57
    cu
    -0.56
    edu
    -0.54
    ixties
    -0.53
    ultan
    -0.53
    POSITIVE LOGITS
    ,,
    0.69
    .,
    0.63
    ,
    0.63
     elusive
    0.62
     stranger
    0.60
    *,
    0.60
    ummer
    0.58
    hea
    0.54
     Disciple
    0.54
    lining
    0.54
    Activations Density 0.027%

    No Known Activations