© 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. 15928
    Prev
    Next
    INDEX
    Explanations

    words related to a specific governmental building

    oai_token-act-pair · gpt-3.5-turbo

    references to the Capitol building and its surrounding political context

    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
    gger
    -0.80
    ãģ¦
    -0.71
    hari
    -0.68
    ordan
    -0.67
     Bucc
    -0.67
    earch
    -0.65
    history
    -0.64
    sweet
    -0.64
    said
    -0.64
    erer
    -0.63
    POSITIVE LOGITS
    itol
    1.05
     Capitol
    0.92
     Hill
    0.91
     dome
    0.79
    rament
    0.73
    ratulations
    0.72
     Heights
    0.72
    reon
    0.71
    itates
    0.71
    adium
    0.70
    Activations Density 0.014%

    No Known Activations