© 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. 1405
    Prev
    Next
    INDEX
    Explanations

    the word "ignore" 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_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
    572
    +0.10
    0.5%
    136
    +0.09
    0.4%
    688
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    61
    +0.10
    0.03
    204
    +0.09
    0.02
    736
    +0.09
    0.02
    Negative Logits
    isher
    -0.79
    ray
    -0.78
    ccording
    -0.78
    uster
    -0.77
    iov
    -0.75
    rim
    -0.73
    ternity
    -0.73
    ikuman
    -0.72
    ernels
    -0.70
    gans
    -0.70
    POSITIVE LOGITS
     warnings
    0.86
    ibly
    0.74
    illy
    0.74
     altogether
    0.72
     cues
    0.70
    fulness
    0.69
     NOTICE
    0.68
     pleas
    0.66
     inconvenient
    0.64
     objections
    0.62
    Activations Density 0.106%

    No Known Activations