© 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. 3-RES-JB
    11. 9132
    Prev
    Next
    INDEX
    Explanations

    metadata or structured data, like lists, dates, and names in text

    oai_token-act-pair · gpt-3.5-turbo

    technical terms and references related to software and computer systems

    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
    Redditor
    -0.71
    ``
    -0.61
     gent
    -0.59
     nonviolent
    -0.56
     tires
    -0.53
     hom
    -0.53
     behav
    -0.51
     pse
    -0.51
     pedoph
    -0.51
     fert
    -0.50
    POSITIVE LOGITS
     Ibid
    0.68
    itsch
    0.67
    ournals
    0.59
     Wraith
    0.58
    owship
    0.57
    ohyd
    0.56
    ocalypse
    0.55
    veyard
    0.52
    rocal
    0.52
    hall
    0.52
    Activations Density 0.620%

    No Known Activations