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

    Neuronpedia

    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. 4412
    Prev
    Next
    INDEX
    Explanations

    contractions and possessives 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
    419
    +0.10
    0.4%
    196
    +0.09
    0.4%
    2
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    512
    +0.10
    0.09
    540
    +0.09
    0.08
    55
    +0.09
    0.07
    Negative Logits
    enne
    -0.67
    ensis
    -0.67
    代
    -0.66
    reen
    -0.65
    rin
    -0.63
    estones
    -0.61
     lasts
    -0.61
    rival
    -0.60
    atial
    -0.59
     satell
    -0.59
    POSITIVE LOGITS
     unfortunate
    0.81
     why
    0.74
     Helpful
    0.73
     frustrating
    0.72
     speculation
    0.71
     counterproductive
    0.69
    laugh
    0.69
     moot
    0.68
    amental
    0.67
     frustration
    0.66
    Activations Density 0.376%

    No Known Activations