© 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. 7189
    Prev
    Next
    INDEX
    Explanations

    words related to villains, especially when they are described as dangerous or menacing

    oai_token-act-pair · gpt-3.5-turbo

    references to villains in narratives

    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
    ollen
    -0.82
    galitarian
    -0.81
    independent
    -0.79
    issance
    -0.74
    sterdam
    -0.73
    press
    -0.71
    choes
    -0.71
    obar
    -0.71
    apers
    -0.69
    rique
    -0.68
    POSITIVE LOGITS
     villain
    1.14
    ous
    1.07
     villains
    1.02
    ously
    0.90
     Bane
    0.86
     mastermind
    0.85
     antagonist
    0.76
    esses
    0.73
     satir
    0.71
    CVE
    0.68
    Activations Density 0.013%

    No Known Activations