© 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. 11-RES-JB
    11. 13164
    Prev
    Next
    INDEX
    Explanations

    subreddit indicators denoted by '/r/' followed by alphanumeric characters

    oai_token-act-pair · gpt-3.5-turbo

    references to subreddit communities and their associated topics

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 11-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    11
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     SOS
    -0.69
     Corpus
    -0.66
     curs
    -0.65
     vein
    -0.65
     CPR
    -0.61
     GPS
    -0.60
     Fernand
    -0.60
     Ore
    -0.59
     Tsarnaev
    -0.58
     expend
    -0.58
    POSITIVE LOGITS
    outine
    1.40
    abbit
    1.34
    ussia
    1.32
    aspberry
    1.32
    uling
    1.29
    ifle
    1.27
    uler
    1.27
    outing
    1.26
    ifles
    1.25
    uled
    1.25
    Activations Density 0.015%

    No Known Activations