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. 43364
    Prev
    Next
    INDEX
    Explanations

    phrases indicating precarious or critical 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
    608
    +0.17
    0.8%
    90
    +0.11
    0.5%
    207
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    608
    +0.17
    0.03
    90
    +0.11
    0.02
    207
    +0.10
    0.02
    Negative Logits
    raq
    -0.75
    nice
    -0.70
    UFF
    -0.68
     GOODMAN
    -0.65
    note
    -0.65
     Perspective
    -0.64
    inet
    -0.62
     Brief
    -0.62
     perspect
    -0.62
    isms
    -0.61
    POSITIVE LOGITS
     extinction
    1.09
     becoming
    1.00
     bankruptcy
    0.99
     starvation
    0.97
     completing
    0.96
     annihilation
    0.92
     collapse
    0.92
     elimination
    0.91
     meltdown
    0.91
     demolition
    0.90
    Activations Density 0.046%

    No Known Activations