© 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. 1-RES-JB
    11. 6607
    Prev
    Next
    INDEX
    Explanations

    words related to failure or inability to succeed or accomplish a task

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "fail" in various contexts

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 1-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    1
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    enfranch
    -0.79
    til
    -0.74
    ourt
    -0.70
    EStream
    -0.70
    itution
    -0.66
    selves
    -0.66
     Pione
    -0.66
    iliary
    -0.65
    collar
    -0.65
    tip
    -0.64
    POSITIVE LOGITS
     fail
    1.14
     miser
    1.11
    fail
    1.05
     failure
    0.99
     failures
    0.90
     fails
    0.87
    lessly
    0.85
     horribly
    0.83
     catast
    0.82
    Failure
    0.81
    Activations Density 0.013%

    No Known Activations