© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    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. 12968
    Prev
    Next
    INDEX
    Explanations

    references to unfairness, unjust treatment, and deceptive practices

    oai_token-act-pair · gpt-3.5-turboTriggered 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
    240
    +0.10
    0.4%
    464
    +0.09
    0.4%
    251
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    140
    +0.10
    0.02
    335
    +0.09
    0.02
    251
    +0.09
    0.01
    Negative Logits
    hent
    -0.91
    gdala
    -0.80
    apse
    -0.78
    udder
    -0.76
    lang
    -0.72
    ince
    -0.71
    audi
    -0.69
    natureconservancy
    -0.69
     Fn
    -0.68
    apses
    -0.67
    POSITIVE LOGITS
    nesses
    0.98
     advantage
    0.95
     Advantage
    0.95
     unfair
    0.89
     disadvantage
    0.88
     burden
    0.86
     burdens
    0.82
     dismissal
    0.82
     disadvant
    0.81
     advantages
    0.80
    Activations Density 0.035%

    No Known Activations