© 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 EvalsExportsinterp-engineNEWAPI 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. 8-RES-JB
    11. 11744
    Prev
    Next
    INDEX
    Explanations

    legal, political, and governmental terms and entities

    oai_token-act-pair · gpt-3.5-turbo

    references to political and social organizations or entities

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    •Skip Features
    the skipgrams

    No Comments

    Negative Logits
    enegger
    -0.75
    terday
    -0.61
    abwe
    -0.61
     Rica
    -0.60
    theless
    -0.59
     advoc
    -0.59
    renheit
    -0.58
     tradem
    -0.58
     conclud
    -0.55
    itored
    -0.55
    POSITIVE LOGITS
    iest
    0.70
     genus
    0.68
     portion
    0.67
    osphere
    0.66
     axis
    0.65
     hypothesis
    0.61
     element
    0.60
     bourgeoisie
    0.60
     subreddit
    0.59
     version
    0.58
    Activations Density 0.983%

    No Known Activations