© 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. 22344
    Prev
    Next
    INDEX
    Explanations

    job titles and positions of authority

    oai_token-act-pair · gpt-4o-miniTriggered 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
    430
    +0.17
    0.7%
    496
    +0.12
    0.5%
    320
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    264
    +0.17
    0.04
    371
    +0.12
    0.05
    401
    +0.11
    0.04
    Negative Logits
     reper
    -0.79
     peril
    -0.78
     vain
    -0.78
     needles
    -0.77
     queens
    -0.76
     queen
    -0.76
     punct
    -0.75
     machines
    -0.73
     dwar
    -0.73
     foss
    -0.73
    POSITIVE LOGITS
     Manager
    1.24
     Candidate
    1.19
     Officer
    1.17
     Coordinator
    1.17
     Partner
    1.10
     Producer
    1.08
     Advisor
    1.08
     Director
    1.07
     Ambassador
    1.05
     Executive
    1.05
    Activations Density 0.126%

    No Known Activations