© 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. 0-RES-JB
    11. 14662
    Prev
    Next
    INDEX
    Explanations

    superlatives indicating extremity or negativity

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "worst" used in various contexts to describe negative situations or experiences

    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
    uments
    -0.71
    FTWARE
    -0.67
    pher
    -0.64
    smith
    -0.64
    SHIP
    -0.64
    ée
    -0.63
    ption
    -0.59
    ette
    -0.59
    adel
    -0.58
    ça
    -0.58
    POSITIVE LOGITS
     worst
    3.58
    worst
    2.94
     Worst
    2.53
     worse
    2.02
     deadliest
    1.83
     darkest
    1.81
     hardest
    1.74
    orst
    1.73
     weakest
    1.66
     heaviest
    1.64
    Activations Density 0.011%

    No Known Activations