© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 5048
    Prev
    Next
    INDEX
    Explanations

    information related to crime incidents, especially shootings and homicides

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    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
    640
    +0.28
    1.3%
    442
    +0.13
    0.6%
    160
    +0.12
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    640
    +0.28
    0.11
    442
    +0.13
    0.10
    160
    +0.12
    0.10
    Negative Logits
     favour
    -0.97
     persuasion
    -0.93
     commissions
    -0.87
     optim
    -0.87
     striving
    -0.86
     innov
    -0.86
     appoint
    -0.85
     renegoti
    -0.84
     aspire
    -0.84
     enriched
    -0.84
    POSITIVE LOGITS
    Investigators
    1.77
    Authorities
    1.73
    Police
    1.72
    Officers
    1.62
    Officials
    1.56
    According
    1.52
    Detect
    1.41
    Copyright
    1.40
    Witness
    1.40
    Anyone
    1.39
    Activations Density 0.308%

    No Known Activations