© 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. 0-RES-JB
    11. 13137
    Prev
    Next
    INDEX
    Explanations

    commands or instructions starting with "Don't."

    oai_token-act-pair · gpt-3.5-turbo

    phrases that start with "Don't" or "Don’t."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 0-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    0
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     bound
    -0.73
     mounts
    -0.68
     marks
    -0.67
     within
    -0.66
     results
    -0.66
     resolved
    -0.64
     crown
    -0.64
     ens
    -0.64
     pleased
    -0.63
     thus
    -0.62
    POSITIVE LOGITS
    Don
    3.34
    DON
    2.30
     Don
    2.14
    don
    2.13
     DON
    1.76
    Do
    1.46
    Stay
    1.30
    Give
    1.29
    Tell
    1.29
    Never
    1.28
    Activations Density 0.016%

    No Known Activations