Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 1536
    5. 8-RES_FS1536-JB
    6. 517
    Prev
    Next
    INDEX
    Explanations

    mention or discussion of hypothetical scenarios and comparisons in the context of sports or statistics

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    397
    +0.11
    0.5%
    291
    +0.10
    0.4%
    348
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    397
    +0.11
    0.09
    139
    +0.10
    0.08
    762
    +0.09
    0.10
    Negative Logits
     WWF
    -0.66
     moreover
    -0.58
     heavyweight
    -0.54
    istries
    -0.54
    ertodd
    -0.53
     Ducks
    -0.53
    emate
    -0.53
    annels
    -0.52
     dou
    -0.51
     Griffin
    -0.51
    POSITIVE LOGITS
     wont
    0.78
     hitherto
    0.74
     accustomed
    0.72
    Downloadha
    0.70
    ":[
    0.69
     traditionally
    0.65
    ault
    0.65
     previously
    0.64
     [|
    0.64
     Tome
    0.63
    Activations Density 2.074%

    No Known Activations