© 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. 2-RES-JB
    11. 22306
    Prev
    Next
    INDEX
    Explanations

    words related to negative qualities or actions, particularly those associated with deceit or harm

    oai_token-act-pair · gpt-3.5-turbo

    recurrent mentions of the word "wicked" and its variations

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 2-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    2
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    anwhile
    -0.89
    ADRA
    -0.87
    ablish
    -0.87
    uther
    -0.86
    ĸļ
    -0.86
    pora
    -0.85
    ylum
    -0.84
    PsyNetMessage
    -0.82
    Downloadha
    -0.80
    minent
    -0.78
    POSITIVE LOGITS
    nesses
    1.05
    ness
    0.84
    ened
    0.79
     Wicked
    0.74
     wed
    0.74
     wicked
    0.72
    itous
    0.70
     stra
    0.69
    est
    0.69
     stripes
    0.68
    Activations Density 0.017%

    No Known Activations