© 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. 10342
    Prev
    Next
    INDEX
    Explanations

    phrases related to propaganda and misinformation

    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
    447
    +0.12
    0.5%
    288
    +0.11
    0.5%
    140
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.12
    0.05
    291
    +0.11
    0.04
    140
    +0.10
    0.04
    Negative Logits
    aird
    -0.91
    angan
    -0.69
    aldo
    -0.68
    arkable
    -0.68
    iosyn
    -0.67
    itaire
    -0.67
    irements
    -0.65
    anse
    -0.64
    riott
    -0.64
     Eucl
    -0.64
    POSITIVE LOGITS
     propag
    1.07
     propaganda
    1.04
     disinformation
    1.03
     tactics
    1.02
     dissemin
    1.00
     pedd
    0.97
     tactic
    0.92
    aganda
    0.89
     leaflets
    0.88
     perpetrated
    0.87
    Activations Density 0.207%

    No Known Activations