© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports 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. 8077
    Prev
    Next
    INDEX
    Explanations

    words related to negative emotions or reactions such as being appalled or horrified

    oai_token-act-pair · gpt-3.5-turbo

    expressions of strong emotional responses, particularly negative sentiments like being appalled or horrified

    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
    opic
    -0.73
    grade
    -0.70
    build
    -0.69
    inth
    -0.68
     installment
    -0.68
    graded
    -0.66
    ingham
    -0.66
    ultimate
    -0.66
    estyles
    -0.65
    enture
    -0.65
    POSITIVE LOGITS
     aback
    1.10
     aroused
    1.00
     horrified
    0.97
     amused
    0.95
     appalled
    0.93
     disgusted
    0.92
     stunned
    0.91
     shocked
    0.89
     irritated
    0.88
     fascinated
    0.88
    Activations Density 0.035%

    No Known Activations