Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 4-RES-JB
    6. 19319
    Prev
    Next
    INDEX
    Explanations

    phrases related to a specific format or marker likely used in a structured document, signalling a potential break or transition

    oai_token-act-pair · gpt-3.5-turbo

    dialogue or conversational exchanges between characters

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

    No Comments

    Negative Logits
     internationally
    -0.76
     estranged
    -0.75
    ividual
    -0.73
     appoint
    -0.72
    osponsors
    -0.71
     geographically
    -0.69
     Burnett
    -0.69
     divest
    -0.68
     orchestr
    -0.68
     builder
    -0.66
    POSITIVE LOGITS
    Chapter
    1.18
    Suddenly
    1.09
    âĶĢâĶĢâĶĢâĶĢ
    1.01
    Fuck
    1.00
    CHAPTER
    1.00
    Eventually
    0.96
    Then
    0.96
    ³³³³
    0.96
    Advertisements
    0.95
    Thankfully
    0.95
    Activations Density 0.306%

    No Known Activations