© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI 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. 3-RES-JB
    11. 23644
    Prev
    Next
    INDEX
    Explanations

    instances of the word "undercover."

    oai_token-act-pair · gpt-3.5-turbo

    references to undercover activities and operations

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 3-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    3
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    È
    -0.79
    gran
    -0.79
    RR
    -0.71
    grave
    -0.69
     Preferred
    -0.67
    plex
    -0.67
    itsch
    -0.67
    cing
    -0.67
    ð
    -0.66
    Stone
    -0.66
    POSITIVE LOGITS
     undercover
    1.23
     informant
    1.03
     mole
    0.91
     sting
    0.87
     infiltr
    0.86
     informants
    0.83
     detective
    0.78
     investigator
    0.75
     uncover
    0.75
     spying
    0.72
    Activations Density 0.012%

    No Known Activations