© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports 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. 206
    Prev
    Next
    INDEX
    Explanations

    terms related to LGBTQ+ issues, specifically focusing on the word "Hom" followed by a number indicating the strength of activation

    oai_token-act-pair · gpt-3.5-turbo

    references to homosexuality and related terms

    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
    UGE
    -0.75
    utical
    -0.70
    ãģį
    -0.67
     shroud
    -0.67
    FUL
    -0.65
    CLASSIFIED
    -0.65
     sender
    -0.64
    BuyableInstoreAndOnline
    -0.64
    IBLE
    -0.64
    hower
    -0.63
    POSITIVE LOGITS
    estead
    1.40
    emade
    1.39
    ework
    1.21
    eless
    1.18
    osexual
    1.13
    ocaust
    1.07
    neys
    1.07
    icide
    1.05
    ogeneous
    1.04
    ogeneity
    0.98
    Activations Density 0.005%

    No Known Activations