© 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 768
    9. 8-RES_FS768-JB
    11. 724
    Prev
    Next
    INDEX
    Explanations

    questions and statements related to inquiry and criticism

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    218
    +0.13
    0.6%
    662
    +0.10
    0.4%
    37
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    287
    +0.13
    0.09
    449
    +0.10
    0.08
    60
    +0.10
    0.07
    Negative Logits
     Dragonbound
    -0.70
    perty
    -0.68
    displayText
    -0.65
    isphere
    -0.63
    ãģ«
    -0.62
    RAFT
    -0.61
    verage
    -0.60
    cko
    -0.60
    swick
    -0.58
    ãģĮ
    -0.58
    POSITIVE LOGITS
    olated
    0.91
    olation
    0.82
    berra
    0.80
     anybody
    0.74
    olate
    0.74
     anyone
    0.74
    fred
    0.73
    ocy
    0.68
    peria
    0.67
    abella
    0.65
    Activations Density 2.175%

    No Known Activations