© 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. 26764
    Prev
    Next
    INDEX
    Explanations

    instances of the word "listen" and its various forms, indicating a focus on auditory engagement

    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
    759
    +0.12
    0.5%
    541
    +0.11
    0.5%
    95
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    70
    +0.12
    0.02
    759
    +0.11
    0.02
    95
    +0.11
    0.02
    Negative Logits
    mber
    -0.71
     Janeiro
    -0.69
    esan
    -0.66
    rament
    -0.65
     Hels
    -0.64
    hire
    -0.64
    \/\/
    -0.63
    enture
    -0.61
    involved
    -0.60
     Palm
    -0.60
    POSITIVE LOGITS
     attent
    1.22
     aloud
    0.96
     listened
    0.92
     passively
    0.90
     closely
    0.88
     listening
    0.87
     listens
    0.84
     listen
    0.83
     loudly
    0.83
     comprehension
    0.80
    Activations Density 0.010%

    No Known Activations