© 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 49152
    9. 8-RES_FS49152-JB
    11. 26658
    Prev
    Next
    INDEX
    Explanations

    references to technical details and explanations related to software development

    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
    497
    +0.12
    0.6%
    459
    +0.11
    0.5%
    655
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    60
    +0.12
    0.03
    97
    +0.11
    0.02
    642
    +0.10
    0.02
    Negative Logits
     UNCLASSIFIED
    -0.80
    ocaust
    -0.73
    irlfriend
    -0.70
    pires
    -0.70
    onna
    -0.69
    pired
    -0.68
    ometown
    -0.68
    usalem
    -0.68
    uclear
    -0.67
    asketball
    -0.65
    POSITIVE LOGITS
     pitfalls
    1.05
     misconceptions
    0.93
     challeng
    0.90
     concepts
    0.89
     syntax
    0.85
     terminology
    0.84
     modifications
    0.82
     definitions
    0.82
     disadvantages
    0.82
     usability
    0.81
    Activations Density 0.238%

    No Known Activations