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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI 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. 8-RES-JB
    11. 12004
    Prev
    Next
    INDEX
    Explanations

    phrases indicating the best or only method to achieve a particular outcome

    oai_token-act-pair · gpt-3.5-turbo

    phrases that express methods or approaches to achieve goals

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    usters
    -0.70
    psc
    -0.66
    ancies
    -0.65
    asts
    -0.61
    Kit
    -0.59
    igrated
    -0.59
    Shape
    -0.58
    origin
    -0.57
    ivered
    -0.57
    drawn
    -0.57
    POSITIVE LOGITS
    ":["
    0.65
     anybody
    0.63
    soever
    0.62
    İ
    0.61
     anyone
    0.61
    ever
    0.60
    aucus
    0.60
     you
    0.58
    ":"
    0.57
    allo
    0.56
    Activations Density 0.084%

    No Known Activations