© 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 49152
    9. 8-RES_FS49152-JB
    11. 37177
    Prev
    Next
    INDEX
    Explanations

    phrases that begin with the word "don’t" or variations of it

    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
    152
    +0.11
    0.5%
    696
    +0.10
    0.5%
    373
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    696
    +0.11
    0.04
    152
    +0.10
    0.03
    360
    +0.10
    0.03
    Negative Logits
     pus
    -0.66
     DRAGON
    -0.64
    EStreamFrame
    -0.63
    EStream
    -0.62
    ħĭ
    -0.61
     Species
    -0.60
     spoiled
    -0.60
    phal
    -0.59
     Featured
    -0.59
    milo
    -0.59
    POSITIVE LOGITS
    't
    1.54
    ations
    0.92
    ned
    0.91
    ately
    0.91
    atives
    0.89
    ÃŃ
    0.87
    nel
    0.87
    ning
    0.87
    ovan
    0.87
    itely
    0.86
    Activations Density 0.023%

    No Known Activations