© 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 · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    5. GPT2-Small
    7. Residual Stream
    9. 3-RES-JB
    11. 3026
    Prev
    Next
    INDEX
    Explanations

    words related to intense interest or fixation on a particular topic

    oai_token-act-pair · gpt-3.5-turbo

    terms related to obsession and addiction

    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
    ND
    -0.77
    channelAvailability
    -0.68
     withheld
    -0.67
    Lake
    -0.65
    DragonMagazine
    -0.65
    Du
    -0.62
     dated
    -0.61
    University
    -0.61
    IFF
    -0.61
    NA
    -0.61
    POSITIVE LOGITS
    adoes
    0.82
    ados
    0.79
    kson
    0.76
    ulously
    0.74
    ules
    0.74
    âĶģ
    0.71
    idious
    0.71
     obsessed
    0.68
    otti
    0.68
    ness
    0.68
    Activations Density 0.028%

    No Known Activations