© 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. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    5. GPT2-Small
    7. Residual Post - 32k
    9. 8-RES_POST_32K-OAI
    11. 5083
    Prev
    Next
    INDEX
    Explanations

    punctuation marks, particularly commas, in the text

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.8.hook_resid_post
    Hook Layer
    8
    Architecture
    standard
    Context Size
    64
    Dataset
    Skylion007/openwebtext
    Activation Function
    topk
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    751
    +0.12
    0.5%
    165
    +0.10
    0.4%
    404
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    43
    +0.12
    0.09
    546
    +0.10
    0.09
    297
    +0.10
    0.09
    Negative Logits
    aez
    -0.71
    bucks
    -0.66
    ======
    -0.62
     Abstract
    -0.62
    stract
    -0.61
    amoto
    -0.60
    onym
    -0.59
    TEXT
    -0.58
    ertodd
    -0.58
    illas
    -0.56
    POSITIVE LOGITS
     whose
    1.43
     which
    1.38
    whose
    1.35
    which
    1.21
     whom
    1.18
     where
    1.01
     who
    0.95
     whence
    0.91
    where
    0.88
    who
    0.85
    Activations Density 0.839%

    No Known Activations