Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 49152
    5. 8-RES_FS49152-JB
    6. 25571
    Prev
    Next
    INDEX
    Explanations

    phrases related to configuration and setup processes in 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
    373
    +0.14
    0.6%
    610
    +0.11
    0.5%
    732
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    326
    +0.14
    0.03
    727
    +0.11
    0.02
    610
    +0.10
    0.02
    Negative Logits
    oustic
    -0.72
     dazz
    -0.72
     discriminate
    -0.72
     stunts
    -0.69
     torped
    -0.69
     airs
    -0.69
     misunderstand
    -0.68
     discriminating
    -0.68
     cruise
    -0.68
     jugg
    -0.68
    POSITIVE LOGITS
     ~/.
    1.49
     folder
    1.42
     ~/
    1.33
    folder
    1.31
     directory
    1.22
     file
    1.17
     folders
    1.17
    directory
    1.16
    :\
    1.16
     '/
    1.15
    Activations Density 0.113%

    No Known Activations