© 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
    2. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    3. GPT2-Small
    4. Residual Post - 32k
    5. 10-RES_POST_32K-OAI
    6. 28741
    Prev
    Next
    INDEX
    Explanations

    phrases that refer to exceptions or negative statements

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.10.hook_resid_post
    Hook Layer
    10
    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
    469
    +0.17
    0.8%
    735
    +0.10
    0.5%
    438
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    735
    +0.17
    0.01
    97
    +0.10
    0.01
    469
    +0.10
    0.01
    Negative Logits
    ICA
    -0.81
    urrent
    -0.81
    ERAL
    -0.78
    TOR
    -0.78
    ESE
    -0.76
    ulton
    -0.74
    ICO
    -0.72
    ouses
    -0.72
    lambda
    -0.70
    INO
    -0.70
    POSITIVE LOGITS
     Helpful
    1.24
     Buy
    0.68
     circumcision
    0.68
     Hebrew
    0.67
     Narr
    0.62
     Psal
    0.61
     Abraham
    0.61
     Practices
    0.59
     Words
    0.59
     Verse
    0.58
    Activations Density 0.073%

    No Known Activations