© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI 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. 1314
    Prev
    Next
    INDEX
    Explanations

    comparisons between different entities or individuals

    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
    396
    +0.12
    0.5%
    469
    +0.11
    0.5%
    669
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    396
    +0.12
    0.04
    539
    +0.11
    0.03
    692
    +0.10
    0.03
    Negative Logits
    oother
    -0.83
    imity
    -0.78
    EMENT
    -0.78
    eeper
    -0.75
    ourt
    -0.73
    raine
    -0.73
    antage
    -0.71
    nr
    -0.69
    enario
    -0.66
    )].
    -0.65
    POSITIVE LOGITS
     usual
    1.13
     ever
    1.10
     anything
    0.98
     EVER
    0.87
     any
    0.81
     anybody
    0.79
     average
    0.78
     anyone
    0.78
    average
    0.76
     anywhere
    0.75
    Activations Density 0.066%

    No Known Activations