© 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
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 3-RES-JB
    6. 6243
    Prev
    Next
    INDEX
    Explanations

    maze-like structures or situations

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "maze" and its variations, indicating a focus on complex structures or challenges

    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
    izations
    -0.99
    erd
    -0.81
    isations
    -0.81
    umption
    -0.78
    izers
    -0.69
    ded
    -0.69
    Ķ
    -0.69
    boards
    -0.68
    arer
    -0.68
    ization
    -0.67
    POSITIVE LOGITS
     maze
    0.99
     Maze
    0.93
     labyrinth
    0.78
    yrinth
    0.76
    ĸļ
    0.73
     Mania
    0.73
    ingly
    0.71
    uzzle
    0.70
     Ples
    0.68
     Malfoy
    0.68
    Activations Density 0.021%

    No Known Activations