© 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. 3-RES-JB
    11. 16777
    Prev
    Next
    INDEX
    Explanations

    phrases related to legal and judicial proceedings, particularly mentioning judges and officers

    oai_token-act-pair · gpt-3.5-turbo

    references to judges and officers involved in legal or governmental contexts

    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
     Territories
    -0.86
    Hur
    -0.73
     Thing
    -0.70
     Virus
    -0.69
     Shore
    -0.69
     Rover
    -0.66
    LAN
    -0.65
     Kind
    -0.65
     Coast
    -0.65
     Sabha
    -0.64
    POSITIVE LOGITS
    hip
    1.56
    hips
    1.40
    paces
    1.20
    ervatives
    1.15
    aurus
    1.10
    ettings
    1.06
    mith
    1.01
    poons
    0.97
    heet
    0.96
    ervative
    0.95
    Activations Density 0.097%

    No Known Activations