© 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 24576
    9. 8-RES_FS24576-JB
    11. 19870
    Prev
    Next
    INDEX
    Explanations

    names or terms specifically containing the sequence "Ka" followed by a single non-zero activation value

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    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
    31
    +0.12
    0.5%
    621
    +0.10
    0.5%
    683
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    659
    +0.12
    0.02
    343
    +0.10
    0.02
    201
    +0.10
    0.01
    Negative Logits
    wcsstore
    -0.73
    ROR
    -0.73
    Ö¼
    -0.70
    OWER
    -0.69
    service
    -0.66
    cream
    -0.66
    çīĪ
    -0.66
    ï¸ı
    -0.65
    lessly
    -0.63
    ently
    -0.63
    POSITIVE LOGITS
    iser
    1.03
    ichi
    1.00
    pler
    0.99
    oru
    0.92
    isers
    0.91
    plan
    0.89
    iba
    0.87
    osta
    0.87
    plin
    0.86
    istically
    0.86
    Activations Density 0.025%

    No Known Activations