© 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 · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    5. GPT2-Small
    7. Residual Stream
    9. 3-RES-JB
    11. 17746
    Prev
    Next
    INDEX
    Explanations

    Islamic or Middle Eastern names, particularly "Ali"

    oai_token-act-pair · gpt-3.5-turbo

    references to the name "Ali."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 3-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.hook_resid_pre
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.3.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    3
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    olicy
    -0.73
    IAL
    -0.70
    terson
    -0.69
    acters
    -0.67
    essee
    -0.67
    idges
    -0.66
    manship
    -0.66
    umbn
    -0.66
     Cosponsors
    -0.65
    deck
    -0.65
    POSITIVE LOGITS
    ensis
    1.07
    yah
    1.03
    qi
    0.91
    ña
    0.90
     Lama
    0.88
    ali
    0.88
    ë
    0.87
    ñ
    0.86
    Äĩ
    0.82
    asing
    0.82
    Activations Density 0.013%

    No Known Activations