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

    Neuronpedia

    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. 4-RES-JB
    11. 16237
    Prev
    Next
    INDEX
    Explanations

    comparative statements or expressions indicating improvement

    oai_token-act-pair · gpt-3.5-turbo

    phrases expressing the concept of improvement or being "better."

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

    No Comments

    Negative Logits
     Pione
    -0.69
    Py
    -0.64
     Dresden
    -0.64
    aly
    -0.63
    idon
    -0.63
     Straw
    -0.62
    FK
    -0.61
    gemony
    -0.60
    mberg
    -0.60
    leted
    -0.60
    POSITIVE LOGITS
     suited
    1.09
     than
    1.09
     behaved
    0.96
     acquainted
    0.84
     Than
    0.80
     predictor
    0.76
    seller
    0.75
     manag
    0.75
     vers
    0.74
     quality
    0.74
    Activations Density 0.033%

    No Known Activations