© 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. 10-RES-JB
    11. 10449
    Prev
    Next
    INDEX
    Explanations

    words related to physical violence, such as "beaten", "stabbed", "raped", and "attacked"

    oai_token-act-pair · gpt-3.5-turbo

    instances of violence and physical assault

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    10
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     understatement
    -0.70
    rium
    -0.69
    gravity
    -0.67
    20439
    -0.66
    Balt
    -0.66
    orers
    -0.64
    orders
    -0.64
     Starr
    -0.64
    aird
    -0.63
    natureconservancy
    -0.63
    POSITIVE LOGITS
     senseless
    1.08
     violently
    1.07
     merciless
    1.05
     unconscious
    1.03
     brutally
    0.99
     by
    0.95
     repeatedly
    0.88
     throats
    0.86
     unarmed
    0.85
     verbally
    0.79
    Activations Density 0.126%

    No Known Activations