© 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. 3309
    Prev
    Next
    INDEX
    Explanations

    words related to compensating for something, typically in a negative context like losses, mistakes, or harm

    oai_token-act-pair · gpt-3.5-turbo

    occurrences of the word "compensate" and its variations

    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
     Willow
    -0.76
     Dresden
    -0.73
     Blazers
    -0.69
     Raiders
    -0.69
    street
    -0.69
    nen
    -0.68
     Deer
    -0.68
    stead
    -0.67
     Devils
    -0.67
    328
    -0.67
    POSITIVE LOGITS
     compensate
    1.23
     compensated
    1.14
     compens
    1.03
    ments
    1.02
    MENTS
    0.86
     compensation
    0.82
     offsets
    0.79
     adequately
    0.78
     ventilation
    0.74
    ustration
    0.74
    Activations Density 0.008%

    No Known Activations