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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & 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. 37082
    Prev
    Next
    INDEX
    Explanations

    phrases emphasizing the concept of being exclusively or solely about something

    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
    441
    +0.12
    0.5%
    99
    +0.10
    0.5%
    106
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    441
    +0.12
    0.01
    106
    +0.10
    0.01
    297
    +0.10
    0.01
    Negative Logits
    orrow
    -0.72
     Cards
    -0.67
    lain
    -0.67
    asta
    -0.66
     Everywhere
    -0.65
    older
    -0.65
    ocker
    -0.65
     Already
    -0.65
     Seeds
    -0.63
     Syndrome
    -0.63
    POSITIVE LOGITS
     cosmetic
    0.96
     utilitarian
    0.91
     speculative
    0.86
    technical
    0.84
    rative
    0.84
     aesthetic
    0.83
     circumst
    0.80
     subjective
    0.80
     superficial
    0.80
     pragmatic
    0.79
    Activations Density 0.007%

    No Known Activations