© 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. 6300
    Prev
    Next
    INDEX
    Explanations

    terms related to the concept of 'sinking' or 'collapse'

    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
    99
    +0.10
    0.4%
    241
    +0.09
    0.4%
    507
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.10
    0.02
    298
    +0.09
    0.01
    271
    +0.09
    0.01
    Negative Logits
    kson
    -0.75
     SX
    -0.66
    ============
    -0.63
    BOOK
    -0.61
    ³³
    -0.60
     Dragonbound
    -0.58
     incorrect
    -0.58
    ioch
    -0.57
     Sharif
    -0.56
     calendars
    -0.55
    POSITIVE LOGITS
    hole
    0.94
    mit
    0.90
    ucer
    0.86
    minist
    0.85
    itate
    0.82
    holes
    0.82
    erman
    0.78
    fish
    0.78
    ibur
    0.76
    deep
    0.74
    Activations Density 0.072%

    No Known Activations