© 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. 20489
    Prev
    Next
    INDEX
    Explanations

    mentions of the word "Batman"

    oai_token-act-pair · gpt-3.5-turbo

    occurrences of the word "Bat" and its variations

    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
    mble
    -0.86
    zanne
    -0.76
    perse
    -0.75
    xia
    -0.72
    olulu
    -0.70
    pressed
    -0.68
    jri
    -0.67
    resents
    -0.65
    bye
    -0.64
    ensional
    -0.63
    POSITIVE LOGITS
    bat
    1.03
    Bat
    0.95
     Bat
    0.94
    room
    0.92
    tered
    0.90
    rooms
    0.87
    anas
    0.82
    ting
    0.80
    illus
    0.80
    allion
    0.78
    Activations Density 0.008%

    No Known Activations