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

    Neuronpedia

    APICircuit 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. 407
    Prev
    Next
    INDEX
    Explanations

    the concept of "running" in various contexts

    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
    270
    +0.12
    0.6%
    627
    +0.10
    0.5%
    155
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    326
    +0.12
    0.07
    270
    +0.10
    0.04
    226
    +0.09
    0.05
    Negative Logits
    alam
    -0.69
    ayers
    -0.64
    repre
    -0.63
    grave
    -0.60
    ortium
    -0.60
    omical
    -0.60
    iasco
    -0.59
    urers
    -0.59
    ippy
    -0.59
    Birth
    -0.58
    POSITIVE LOGITS
    swick
    0.98
    escape
    0.97
    aways
    0.95
    gs
    0.92
    dy
    0.85
     af
    0.84
    ways
    0.83
    running
    0.80
    ners
    0.79
    Disney
    0.79
    Activations Density 1.702%

    No Known Activations