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

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 3072
    9. 8-RES_FS3072-JB
    11. 62
    Prev
    Next
    INDEX
    Explanations

    occurrences of specific names and titles, likely related to politics and government

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs3072-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    3,072
    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
    760
    +0.14
    0.6%
    115
    +0.12
    0.5%
    242
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    115
    +0.14
    0.13
    760
    +0.12
    0.08
    242
    +0.11
    0.08
    Negative Logits
    chrome
    -0.77
     ILCS
    -0.69
    usercontent
    -0.66
    requ
    -0.66
     thri
    -0.66
    Torrent
    -0.64
    pmwiki
    -0.64
     somehow
    -0.64
    Ö
    -0.64
     screw
    -0.63
    POSITIVE LOGITS
     conference
    1.24
     keynote
    1.24
     convened
    1.12
     attendees
    1.10
     podium
    1.07
     briefing
    1.04
    posium
    1.04
     unveiling
    1.03
     presentations
    1.03
     televised
    1.02
    Activations Density 2.475%

    No Known Activations