© 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. 0-RES-JB
    11. 536
    Prev
    Next
    INDEX
    Explanations

    verbs in the past tense

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "took"

    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
     hereafter
    -0.58
    ulate
    -0.57
    oken
    -0.55
    Quantity
    -0.54
     academ
    -0.53
     behalf
    -0.52
    ints
    -0.52
    ide
    -0.52
    endment
    -0.51
    ãĥ¼
    -0.50
    POSITIVE LOGITS
     took
    3.01
     gave
    1.81
     takes
    1.80
     went
    1.75
     threw
    1.72
     undertook
    1.70
     began
    1.67
     drew
    1.66
     stole
    1.64
     blew
    1.63
    Activations Density 0.026%

    No Known Activations