© 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. 6-RES-JB
    11. 12052
    Prev
    Next
    INDEX
    Explanations

    words related to geographical locations, especially focusing on specific place names

    oai_token-act-pair · gpt-3.5-turbo

    references to specific figures associated with the name "Prince."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 6-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.6.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.6.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    6
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    places
    -0.67
    inker
    -0.65
    ventions
    -0.63
    Redditor
    -0.62
     lever
    -0.61
    acle
    -0.61
    baugh
    -0.60
    onom
    -0.60
    ocre
    -0.60
     stitching
    -0.60
    POSITIVE LOGITS
    âĦ¢:
    0.87
    ;;;;;;;;;;;;
    0.87
    VII
    0.80
    reciation
    0.77
    ossom
    0.74
    foundland
    0.73
     Islands
    0.73
     remission
    0.69
    ashington
    0.68
    loo
    0.68
    Activations Density 0.056%

    No Known Activations