© 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. 11220
    Prev
    Next
    INDEX
    Explanations

    phrases starting with "Don't" followed by an action or instruction

    oai_token-act-pair · gpt-3.5-turbo

    negative phrases and expressions of doubt or caution

    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
    ħĭ
    -0.91
    Redd
    -0.81
    milo
    -0.80
    ĥ
    -0.74
    ģĸ
    -0.71
    £ı
    -0.71
    ELD
    -0.71
    edged
    -0.69
     shorth
    -0.68
    Reviewer
    -0.67
    POSITIVE LOGITS
    heny
    0.82
     bother
    0.77
    omsday
    0.76
     underestimate
    0.75
     thou
    0.72
    ations
    0.70
    entious
    0.70
    ardless
    0.70
    ibilities
    0.69
     yourselves
    0.69
    Activations Density 0.078%

    No Known Activations