© 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. 4-RES-JB
    11. 7473
    Prev
    Next
    INDEX
    Explanations

    verbs related to reversing actions or outcomes

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "undo" and its variations, focusing on themes of reversal and retraction

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

    No Comments

    Negative Logits
    asio
    -0.87
    ramid
    -0.79
    ategories
    -0.77
    croft
    -0.77
    tek
    -0.77
    colo
    -0.75
    chens
    -0.74
    rikes
    -0.74
    tom
    -0.72
     Flavoring
    -0.71
    POSITIVE LOGITS
     undo
    1.06
     undone
    0.88
     havoc
    0.80
    undo
    0.75
    issance
    0.66
     erase
    0.64
     disarm
    0.64
     popul
    0.64
     detract
    0.64
     dismantling
    0.64
    Activations Density 0.010%

    No Known Activations