© 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. 1-RES-JB
    11. 10396
    Prev
    Next
    INDEX
    Explanations

    words related to breaking or destroying something, often in a negative or forceful manner

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "busted" and its variations, indicating a focus on arrests or exposure of wrongdoing

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 1-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    1
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    minist
    -0.87
    uther
    -0.86
     unden
    -0.78
    isation
    -0.74
    isable
    -0.73
    apist
    -0.71
    umerable
    -0.71
    eway
    -0.69
    activated
    -0.69
    tymology
    -0.68
    POSITIVE LOGITS
     busted
    1.28
     bust
    0.96
    buster
    0.86
     sag
    0.78
     toe
    0.76
     dope
    0.73
    busters
    0.72
     Bust
    0.71
     buck
    0.71
    kok
    0.70
    Activations Density 0.007%

    No Known Activations