© 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. 7-RES-JB
    11. 2669
    Prev
    Next
    INDEX
    Explanations

    sentences where the phrase "That being said" or "That said" is used

    oai_token-act-pair · gpt-3.5-turbo

    the phrase "that said," indicating a transition or contrast in a statement

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 7-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    7
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    elled
    -0.67
    atos
    -0.65
    prus
    -0.60
    urated
    -0.60
    quad
    -0.59
    asus
    -0.58
    Redditor
    -0.57
    called
    -0.57
    "},"
    -0.57
    ãĤµ
    -0.56
    POSITIVE LOGITS
     anecd
    0.79
     caveats
    0.77
    ,.
    0.74
    itivity
    0.72
     respectfully
    0.70
    lihood
    0.69
     respecting
    0.69
    ,
    0.68
     disclaim
    0.68
     notwithstanding
    0.67
    Activations Density 0.038%

    No Known Activations