© 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. 5972
    Prev
    Next
    INDEX
    Explanations

    communication prompts for returning or exchanging items, with a focus on addressing shipping concerns and customer support

    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
    156
    +0.12
    0.6%
    267
    +0.11
    0.5%
    393
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    156
    +0.12
    0.05
    267
    +0.11
    0.04
    429
    +0.10
    0.04
    Negative Logits
    Georg
    -0.69
    wikipedia
    -0.60
    itol
    -0.60
     Sabha
    -0.59
    dust
    -0.59
    bats
    -0.58
    ©¶æ¥µ
    -0.58
    pins
    -0.58
    assembly
    -0.57
     Estimates
    -0.57
    POSITIVE LOGITS
    're
    1.41
     want
    1.17
    've
    1.13
     decide
    1.07
     encount
    1.06
     wanna
    1.06
     wish
    1.03
     prefer
    1.02
     intend
    1.01
     choose
    1.00
    Activations Density 0.079%

    No Known Activations