© 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. 1286
    Prev
    Next
    INDEX
    Explanations

    phrases expressing contradiction or exception

    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
    393
    +0.13
    0.6%
    203
    +0.13
    0.6%
    8
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    203
    +0.13
    0.06
    8
    +0.13
    0.05
    208
    +0.12
    0.04
    Negative Logits
    ILLE
    -0.66
    asks
    -0.66
    meta
    -0.65
    orter
    -0.62
    endar
    -0.61
    enter
    -0.61
    imgur
    -0.60
    wake
    -0.59
    ory
    -0.59
    FACE
    -0.59
    POSITIVE LOGITS
     rather
    1.39
     nevertheless
    1.29
     nonetheless
    1.17
    rather
    1.09
     merely
    1.08
     suffice
    1.02
     Rather
    0.95
     instead
    0.92
     alas
    0.89
     luckily
    0.87
    Activations Density 0.129%

    No Known Activations