© 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 768
    9. 8-RES_FS768-JB
    11. 709
    Prev
    Next
    INDEX
    Explanations

    expressions of disagreement and contrasting viewpoints

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    208
    +0.11
    0.5%
    8
    +0.11
    0.5%
    417
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    598
    +0.11
    0.11
    203
    +0.11
    0.12
    8
    +0.10
    0.12
    Negative Logits
    edu
    -0.68
    naire
    -0.67
    oire
    -0.66
    ãĤµ
    -0.65
    psy
    -0.63
    uto
    -0.63
    amus
    -0.62
    ãģ®
    -0.61
    roy
    -0.61
    guard
    -0.61
    POSITIVE LOGITS
    tons
    1.42
     nevertheless
    1.41
     alas
    1.41
     nonetheless
    1.35
    chery
    1.21
    chers
    1.15
     fortunately
    1.12
     luckily
    1.08
     beware
    1.06
     hey
    0.99
    Activations Density 1.703%

    No Known Activations