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 49152
    5. 8-RES_FS49152-JB
    6. 43208
    Prev
    Next
    INDEX
    Explanations

    phrases indicating negative consequences or reactions related to events or situations

    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
    481
    +0.23
    1.1%
    218
    +0.10
    0.5%
    447
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    98
    +0.23
    0.06
    703
    +0.10
    0.04
    608
    +0.10
    0.05
    Negative Logits
    atography
    -0.75
    initely
    -0.75
    Laughs
    -0.73
    Null
    -0.73
    Anyway
    -0.72
    ometry
    -0.72
     nodd
    -0.70
    ograp
    -0.68
    uum
    -0.68
    gmail
    -0.67
    POSITIVE LOGITS
     scandals
    1.16
     recent
    1.11
     revelations
    1.08
     setbacks
    1.06
     controversies
    1.05
     disasters
    0.98
     defeats
    0.97
     allegations
    0.96
     developments
    0.95
     fallout
    0.94
    Activations Density 0.219%

    No Known Activations