© 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. 15378
    Prev
    Next
    INDEX
    Explanations

    emotionally charged statements reflecting dissatisfaction with the actions or decisions of others

    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
    393
    +0.12
    0.5%
    533
    +0.11
    0.5%
    679
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    203
    +0.12
    0.03
    533
    +0.11
    0.03
    705
    +0.10
    0.03
    Negative Logits
    orage
    -0.80
    aukee
    -0.73
    ead
    -0.67
    ield
    -0.67
    abase
    -0.67
    æĿ
    -0.65
    undo
    -0.64
    esse
    -0.63
    iHUD
    -0.63
    iaz
    -0.62
    POSITIVE LOGITS
     someone
    0.88
     somebody
    0.83
     they
    0.78
    someone
    0.75
     anyone
    0.74
     we
    0.73
    soever
    0.70
     anybody
    0.67
     there
    0.66
     whoever
    0.66
    Activations Density 0.090%

    No Known Activations