© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports 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. 19775
    Prev
    Next
    INDEX
    Explanations

    information about a writing project related to prisoners and Earth First! initiatives

    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
    120
    +0.11
    0.5%
    160
    +0.11
    0.5%
    125
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    751
    +0.11
    0.02
    125
    +0.11
    0.02
    160
    +0.10
    0.02
    Negative Logits
     manif
    -0.71
     footing
    -0.71
     stopp
    -0.66
    orem
    -0.66
    abal
    -0.65
    ction
    -0.64
     recomp
    -0.64
    ymm
    -0.62
    quartered
    -0.62
    ovember
    -0.61
    POSITIVE LOGITS
    Tumblr
    1.25
    Pinterest
    1.16
    Google
    1.14
    LinkedIn
    1.12
    Pocket
    1.01
    Reddit
    1.00
     Tumblr
    0.99
    Twitter
    0.97
    Email
    0.97
     Pinterest
    0.95
    Activations Density 0.030%

    No Known Activations