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. 0-RES-JB
    6. 5850
    Prev
    Next
    INDEX
    Explanations

    mentions of specific terrorist groups like ISIL (Islamic State of Iraq and the Levant) and Daesh

    oai_token-act-pair · gpt-3.5-turbo

    references to terrorist organizations and their activities

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 0-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    0
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Ups
    -0.71
     bunny
    -0.67
     Richards
    -0.66
     pricey
    -0.64
     Mant
    -0.64
     Coffee
    -0.62
     bench
    -0.60
     Mod
    -0.60
     Blue
    -0.60
     Brown
    -0.59
    POSITIVE LOGITS
     ISIL
    3.73
     Daesh
    2.50
     Isis
    1.50
     PKK
    1.47
     Hezbollah
    1.39
     UNHCR
    1.31
     YPG
    1.25
     ISIS
    1.22
     Ankara
    1.19
    Islamic
    1.18
    Activations Density 0.019%

    No Known Activations