© 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 EvalsExportsAPI 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. 4-RES-JB
    11. 983
    Prev
    Next
    INDEX
    Explanations

    common phrases including the conjunction "and."

    oai_token-act-pair · gpt-3.5-turbo

    the conjunction "and" indicating connections or collaborations between multiple subjects or entities

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 4-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    4
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    =]
    -0.76
    uve
    -0.74
    matter
    -0.71
    nt
    -0.69
    ĨĴ
    -0.68
    uggest
    -0.68
    Deal
    -0.66
    vier
    -0.66
    cms
    -0.66
    Contents
    -0.65
    POSITIVE LOGITS
    rogen
    0.95
    rogens
    0.90
     vice
    0.88
     others
    0.84
     Sons
    0.82
     Associates
    0.78
     consequently
    0.76
     colleagues
    0.75
     therefore
    0.75
     other
    0.68
    Activations Density 0.273%

    No Known Activations