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

    Neuronpedia

    Jacobian LensNEW
    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. 12989
    Prev
    Next
    INDEX
    Explanations

    words related to actions or processes being done, particularly those verbs are related to specific actions like ways of speaking

    oai_token-act-pair · gpt-3.5-turbo

    plural forms of words, particularly those ending in 'ays'

    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
     constitu
    -0.83
     Const
    -0.66
     struct
    -0.65
     const
    -0.63
     mortal
    -0.60
     plural
    -0.59
     en
    -0.59
     Grade
    -0.59
     enchantment
    -0.59
     enum
    -0.58
    POSITIVE LOGITS
    ays
    4.70
    aying
    2.36
    ayed
    2.26
    ay
    2.11
    AY
    1.99
    ayers
    1.77
    ayer
    1.69
    ayson
    1.33
    aid
    1.32
    rays
    1.31
    Activations Density 0.007%

    No Known Activations