© 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 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. 23963
    Prev
    Next
    INDEX
    Explanations

    phrases related to specific names and locations

    oai_token-act-pair · gpt-3.5-turbo

    references to specific individuals and locations, particularly "Buckingham Palace" and "Shirley."

    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
    abl
    -0.84
    oster
    -0.82
    oly
    -0.79
    oxin
    -0.79
    ox
    -0.76
    Ò
    -0.76
    igl
    -0.76
    ovan
    -0.74
    otropic
    -0.74
    eb
    -0.74
    POSITIVE LOGITS
     Manson
    0.84
     Parish
    0.83
    issance
    0.82
     readings
    0.78
     conduc
    0.77
     Shirley
    0.75
     melodies
    0.73
     rehearsal
    0.72
     mathemat
    0.71
     princ
    0.71
    Activations Density 0.027%

    No Known Activations