© 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. 0-RES-JB
    11. 9566
    Prev
    Next
    INDEX
    Explanations

    phrases related to loss of various types

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "lost" in various contexts

    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
    imaru
    -0.75
    Express
    -0.74
    Instruct
    -0.72
    spec
    -0.69
    soType
    -0.65
    Spec
    -0.65
    Reddit
    -0.65
    include
    -0.63
    ograph
    -0.63
    haps
    -0.63
    POSITIVE LOGITS
     lost
    3.57
    lost
    2.78
    Lost
    2.17
     Lost
    2.02
     lose
    1.87
     regained
    1.86
     loses
    1.84
     wasted
    1.72
     losing
    1.69
     missed
    1.69
    Activations Density 0.029%

    No Known Activations