© 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 24576
    9. 8-RES_FS24576-JB
    11. 22104
    Prev
    Next
    INDEX
    Explanations

    expressions related to regret and negative feelings about self-harm decisions

    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
    351
    +0.12
    0.6%
    756
    +0.10
    0.5%
    373
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    351
    +0.12
    0.06
    762
    +0.10
    0.09
    283
    +0.10
    0.02
    Negative Logits
     Originally
    -0.77
    anmar
    -0.64
    htaking
    -0.64
    announced
    -0.64
    chell
    -0.63
     Cosponsors
    -0.63
    atri
    -0.62
    north
    -0.62
    inelli
    -0.61
     Tonight
    -0.61
    POSITIVE LOGITS
     mistakes
    0.99
     subconscious
    0.99
     unconsciously
    0.95
     afterwards
    0.90
     anyways
    0.87
     payoff
    0.86
     worthless
    0.84
     eventually
    0.84
     afterward
    0.83
     outweigh
    0.83
    Activations Density 1.016%

    No Known Activations