© 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. 41689
    Prev
    Next
    INDEX
    Explanations

    expressions of well-wishing or encouragement

    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
    88
    +0.11
    0.5%
    575
    +0.10
    0.4%
    705
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    88
    +0.11
    0.01
    446
    +0.10
    0.01
    705
    +0.09
    0.01
    Negative Logits
    anga
    -0.70
    cept
    -0.65
    cious
    -0.64
    raph
    -0.63
    gemony
    -0.61
    leted
    -0.61
    Ĺ
    -0.60
    akespe
    -0.60
    famous
    -0.60
    ificantly
    -0.60
    POSITIVE LOGITS
     charms
    0.77
     luck
    0.75
     guessing
    0.69
    luster
    0.67
     charm
    0.64
    nings
    0.64
     finding
    0.63
    !:
    0.62
    ????????
    0.62
    !
    0.61
    Activations Density 0.007%

    No Known Activations