© 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. 1-RES-JB
    11. 890
    Prev
    Next
    INDEX
    Explanations

    terms related to sports matchups or competitions

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "vs"

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 1-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    1
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    Catal
    -0.69
     liner
    -0.68
    irlf
    -0.63
    ĸļ士
    -0.63
     fracturing
    -0.62
     searched
    -0.62
    raising
    -0.61
    ocene
    -0.61
    --------------------------------------------------------
    -0.60
    ĵ
    -0.59
    POSITIVE LOGITS
    vs
    1.30
    terday
    0.95
    ky
    0.89
    kaya
    0.88
    kies
    0.88
    pect
    0.86
    ki
    0.81
    vp
    0.77
    ync
    0.76
    imir
    0.75
    Activations Density 0.005%

    No Known Activations