© 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 EvalsExportsinterp-engineNEWAPI 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. 22617
    Prev
    Next
    INDEX
    Explanations

    emotional references to regret and remorse

    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
    624
    +0.13
    0.6%
    288
    +0.12
    0.6%
    98
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    624
    +0.13
    0.01
    151
    +0.12
    0.01
    98
    +0.10
    0.01
    Negative Logits
    ĪĴ
    -0.70
     rigs
    -0.68
    uana
    -0.63
    paio
    -0.63
    æ©Ł
    -0.61
    adj
    -0.61
    女
    -0.61
    soDeliveryDate
    -0.60
    anooga
    -0.59
    icles
    -0.58
    POSITIVE LOGITS
    fully
    1.29
    ful
    1.15
    fulness
    1.07
    ting
    1.00
    imaru
    0.90
    faced
    0.89
    ably
    0.88
    rely
    0.87
    lessly
    0.85
    FUL
    0.82
    Activations Density 0.023%

    No Known Activations