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. 19126
    Prev
    Next
    INDEX
    Explanations

    mentions of connections or interactions between Donald Trump's campaign and Russia, including terms like collusion, meddling, and investigation

    oai_token-act-pair · gpt-3.5-turbo

    references to Russia and its involvement in political contexts

    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
    dash
    -0.76
     Dart
    -0.75
    Thom
    -0.75
    ulton
    -0.73
    eme
    -0.72
    }}}
    -0.71
    prototype
    -0.70
    Label
    -0.69
    draw
    -0.68
    jc
    -0.68
    POSITIVE LOGITS
     dossier
    0.92
     meddling
    0.91
     blackmail
    0.91
     collusion
    0.90
     hoax
    0.90
     targeting
    0.87
     spying
    0.87
     intelligence
    0.86
    ghazi
    0.85
     disinformation
    0.83
    Activations Density 0.027%

    No Known Activations