© 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. 9842
    Prev
    Next
    INDEX
    Explanations

    mentions of threats and actions related to violence or conflict

    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
    207
    +0.09
    0.4%
    165
    +0.09
    0.4%
    410
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    665
    +0.09
    0.03
    608
    +0.09
    0.04
    282
    +0.08
    0.03
    Negative Logits
     Parables
    -0.83
    emis
    -0.76
    quickShipAvailable
    -0.74
    Reviewer
    -0.73
     ECO
    -0.71
    swick
    -0.70
    fortable
    -0.70
    CV
    -0.68
    _>
    -0.68
    ortunate
    -0.67
    POSITIVE LOGITS
     wrath
    1.03
     veto
    1.00
     arrest
    0.99
     retribution
    0.97
     derail
    0.97
     blackmail
    0.97
     annihilation
    0.95
     wipe
    0.95
     boycott
    0.94
     sue
    0.93
    Activations Density 0.180%

    No Known Activations