© 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. 2-RES-JB
    11. 11751
    Prev
    Next
    INDEX
    Explanations

    words related to training or being trained

    oai_token-act-pair · gpt-3.5-turbo

    terms related to training or preparation

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 2-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    2
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    FACE
    -0.72
     Meier
    -0.70
    eu
    -0.69
    adra
    -0.65
    sm
    -0.65
    que
    -0.61
    Minecraft
    -0.60
    ãĥİ
    -0.60
     Sack
    -0.60
    510
    -0.59
    POSITIVE LOGITS
     guiActiveUn
    0.90
     instructors
    0.83
     certific
    0.82
    anooga
    0.81
    ĨĴ
    0.80
     trained
    0.78
    anamo
    0.75
     professionals
    0.75
     veter
    0.74
     acad
    0.72
    Activations Density 0.018%

    No Known Activations