© 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. 11-RES-JB
    11. 2846
    Prev
    Next
    INDEX
    Explanations

    verbs related to various actions and states, as well as structures such as "If X was Y" and "What's more, you can get Z."

    oai_token-act-pair · gpt-3.5-turbo

    forms of the verb "to be."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 11-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    11
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    igators
    -0.69
    warn
    -0.67
    intend
    -0.66
     withd
    -0.66
    lees
    -0.65
    ucc
    -0.65
    erk
    -0.63
    aea
    -0.61
    ailed
    -0.61
     Administ
    -0.58
    POSITIVE LOGITS
     indispensable
    1.03
     synonymous
    1.02
     something
    0.97
     paramount
    0.96
     king
    0.95
     everything
    0.92
     everywhere
    0.89
     destiny
    0.89
     integral
    0.86
     foundational
    0.86
    Activations Density 0.339%

    No Known Activations