© 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 EvalsExportsAPI 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. 3-RES-JB
    11. 4874
    Prev
    Next
    INDEX
    Explanations

    phrases related to medicinal terms

    oai_token-act-pair · gpt-3.5-turbo

    references to medicinal properties and benefits, particularly related to cannabis

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 3-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    3
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    zee
    -0.84
    clair
    -0.81
    plin
    -0.80
    illance
    -0.80
    ploy
    -0.80
    ck
    -0.77
    psons
    -0.76
    CTV
    -0.76
    elsen
    -0.73
    holes
    -0.72
    POSITIVE LOGITS
     medicinal
    0.97
     herbs
    0.95
     cannabis
    0.85
     vegetables
    0.84
     marijuana
    0.84
     plants
    0.83
    uana
    0.83
     cannabinoids
    0.81
     substances
    0.80
     herbal
    0.79
    Activations Density 0.028%

    No Known Activations