© 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. 9-RES-JB
    11. 15996
    Prev
    Next
    INDEX
    Explanations

    phrases related to physical harm, danger, or law enforcement actions

    oai_token-act-pair · gpt-3.5-turbo

    phrases indicating imminent danger and serious physical harm

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 9-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    9
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Panama
    -0.75
     portfolios
    -0.75
     Tea
    -0.72
     Compass
    -0.71
     bankers
    -0.71
    cius
    -0.70
     newsletters
    -0.70
     Quartz
    -0.70
     lobbyists
    -0.69
     folders
    -0.69
    POSITIVE LOGITS
     harm
    1.24
    angering
    1.20
     incapac
    1.15
     bodily
    1.14
     injury
    1.14
     injure
    1.05
     harmed
    1.01
     endanger
    0.98
     inflicted
    0.96
    lethal
    0.96
    Activations Density 0.734%

    No Known Activations