© 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. 3954
    Prev
    Next
    INDEX
    Explanations

    phrases indicating a sequence of steps or actions, often starting with the word "First"

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "First."

    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
    vim
    -0.73
    bos
    -0.72
    sav
    -0.68
    sed
    -0.68
    jad
    -0.66
    pher
    -0.66
    phe
    -0.65
    walking
    -0.64
    ties
    -0.64
    ucl
    -0.64
    POSITIVE LOGITS
     responders
    1.07
     baseman
    0.86
     glance
    0.81
     impressions
    0.81
     Steps
    0.79
     Thoughts
    0.76
     Listen
    0.74
     Amendment
    0.74
     blush
    0.71
     Respond
    0.71
    Activations Density 0.037%

    No Known Activations