© 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. 3-RES-JB
    11. 4341
    Prev
    Next
    INDEX
    Explanations

    phrases related to laws, organizations, and regulations

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "and" in various contexts

    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
    :[
    -0.78
    :(
    -0.76
    apter
    -0.71
    Contents
    -0.69
    !:
    -0.68
    DragonMagazine
    -0.67
    éŃĶ
    -0.66
    Ĥİ
    -0.64
    aws
    -0.64
    nces
    -0.63
    POSITIVE LOGITS
     consequently
    1.04
     therefore
    0.98
     hence
    0.97
     vice
    0.95
     thereby
    0.90
     thus
    0.90
     possibly
    0.88
     secondly
    0.87
    rogen
    0.86
    rogens
    0.85
    Activations Density 0.996%

    No Known Activations