© 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 12288
    9. 8-RES_FS12288-JB
    11. 684
    Prev
    Next
    INDEX
    Explanations

    verbs related to actions and impact, especially focusing on societal changes or developments

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    453
    +0.12
    0.5%
    608
    +0.11
    0.5%
    403
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    403
    +0.12
    0.05
    608
    +0.11
    0.07
    172
    +0.11
    0.05
    Negative Logits
    Ha
    -0.77
    inav
    -0.76
    yd
    -0.69
     sidx
    -0.66
    bow
    -0.64
    ologue
    -0.64
    ha
    -0.63
    ALK
    -0.62
     viz
    -0.61
     helicop
    -0.60
    POSITIVE LOGITS
     us
    0.85
     inex
    0.85
     havoc
    0.76
     astronomers
    0.74
     investors
    0.73
     resentment
    0.72
     employers
    0.72
     landlords
    0.71
     many
    0.71
     unsustainable
    0.71
    Activations Density 0.374%

    No Known Activations