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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 0-RES-JB
    6. 6055
    Prev
    Next
    INDEX
    Explanations

    past tense verbs ending in 'ed'

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "hadn't."

    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
     resulting
    -0.55
     Relative
    -0.54
    em
    -0.53
     Ext
    -0.53
     free
    -0.53
     CFR
    -0.52
     Example
    -0.52
    ielding
    -0.51
    ember
    -0.51
     Values
    -0.50
    POSITIVE LOGITS
     hadn
    3.26
     weren
    1.91
     didn
    1.81
     wasn
    1.80
     hasn
    1.78
    didn
    1.74
     haven
    1.69
     couldn
    1.58
     wouldn
    1.54
     didnt
    1.38
    Activations Density 0.011%

    No Known Activations