Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APISteerSAE EvalsBlog/PodcastNEWSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlog/PodcastGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 6-RES-JB
    6. 10623
    Prev
    Next
    INDEX
    Explanations

    instances of the number "three" in the text

    oai_token-act-pair · gpt-3.5-turbo

    nouns related to quantities or groupings

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    the plural noun in "[two/three/four] [plural noun]"

    The word immediately following a spelled-out number

    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 6-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.6.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.6.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    6
    Activation Function
    relu
    Embeds
    IFrame
    Link
    •Interesting
    Fires on the words *after* a spelled-out number

    No Comments

    Negative Logits
    SPONSORED
    -0.73
     Newsletter
    -0.64
     Leth
    -0.63
     Telegram
    -0.62
     Ellis
    -0.62
     Abdel
    -0.60
     Fre
    -0.60
     Stard
    -0.60
    humane
    -0.59
     Fey
    -0.58
    POSITIVE LOGITS
     apiece
    1.40
     totaling
    1.01
    poons
    0.84
    handled
    0.82
    undred
    0.80
    paces
    0.80
    ixt
    0.77
     respectively
    0.76
     consecut
    0.75
    fif
    0.74
    Activations Density 0.252%

    No Known Activations