© 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 49152
    9. 8-RES_FS49152-JB
    11. 26579
    Prev
    Next
    INDEX
    Explanations

    references to nuclear weapons and missile testing activities

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    414
    +0.10
    0.5%
    179
    +0.09
    0.4%
    472
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    335
    +0.10
    0.02
    115
    +0.09
    0.02
    288
    +0.08
    0.02
    Negative Logits
    ĸļ
    -0.96
     Homeless
    -0.79
     LLP
    -0.79
     Wage
    -0.76
     Pant
    -0.71
     Ashe
    -0.71
     Ez
    -0.71
    idav
    -0.70
     Chest
    -0.69
     Sacrament
    -0.68
    POSITIVE LOGITS
     missile
    1.11
     submarine
    1.09
     missiles
    1.07
     submarines
    1.00
     detonated
    0.94
     warheads
    0.92
     rockets
    0.90
     nuclear
    0.89
     DPRK
    0.87
     rocket
    0.87
    Activations Density 0.057%

    No Known Activations