© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 14827
    Prev
    Next
    INDEX
    Explanations

    intense emotions or personal struggles expressed in a reflective manner

    oai_token-act-pair · gpt-3.5-turbo
    Triggered 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
    Neuron Alignment
    Index
    Value
    % of L₁
    208
    +0.11
    0.5%
    620
    +0.11
    0.5%
    360
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    697
    +0.11
    0.05
    360
    +0.11
    0.06
    551
    +0.10
    0.06
    Negative Logits
     Alleg
    -0.65
    itiz
    -0.65
    çĦ
    -0.64
     Reloaded
    -0.64
     Penguin
    -0.62
    soType
    -0.58
     Maiden
    -0.57
     Policies
    -0.57
     Reduced
    -0.56
     spirited
    -0.56
    POSITIVE LOGITS
     necessarily
    1.14
     bother
    1.11
     hesitate
    1.03
     discriminate
    0.98
     deserve
    0.94
     appreciate
    0.94
     know
    0.92
     dare
    0.91
    necess
    0.88
    intend
    0.88
    Activations Density 0.101%

    No Known Activations