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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 24576
    5. 8-RES_FS24576-JB
    6. 23087
    Prev
    Next
    INDEX
    Explanations

    strong emotional responses or reactions from people in social situations

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    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
    87
    +0.29
    1.4%
    544
    +0.08
    0.4%
    217
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    87
    +0.29
    -0.03
    59
    +0.08
    0.04
    217
    +0.08
    0.04
    Negative Logits
     dependency
    -0.79
    catentry
    -0.78
     inheritance
    -0.74
     Decay
    -0.70
     solves
    -0.68
     poisons
    -0.68
     bankrupt
    -0.68
     Suppose
    -0.67
     blackmail
    -0.67
     bankruptcy
    -0.66
    POSITIVE LOGITS
     attendees
    1.24
     applause
    1.17
     cheering
    1.12
     cheers
    1.12
     spectators
    1.08
     chants
    1.05
     podium
    1.04
     chanting
    1.04
     seated
    1.02
     exhib
    0.99
    Activations Density 1.282%

    No Known Activations