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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & 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. 874
    Prev
    Next
    INDEX
    Explanations

    conjunctions in combination with other words

    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
    481
    +0.13
    0.6%
    393
    +0.11
    0.5%
    203
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    203
    +0.13
    0.11
    655
    +0.11
    0.10
    393
    +0.10
    0.12
    Negative Logits
    zar
    -0.71
    Times
    -0.69
    emi
    -0.69
    uta
    -0.66
    ut
    -0.66
    ibble
    -0.65
    hart
    -0.65
    ji
    -0.65
    du
    -0.65
    shift
    -0.64
    POSITIVE LOGITS
     furthermore
    1.45
     consequently
    1.42
     moreover
    1.37
     therefore
    1.30
     hence
    1.26
     secondly
    1.23
     thus
    1.21
     rightfully
    1.12
     although
    1.10
     thence
    1.10
    Activations Density 1.319%

    No Known Activations