© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    5. GPT2-Small
    7. Residual Post - 32k
    9. 10-RES_POST_32K-OAI
    11. 27400
    Prev
    Next
    INDEX
    Explanations

    phrases related to paternity and familial relationships

    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
    184
    +0.12
    0.6%
    601
    +0.10
    0.5%
    518
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    139
    +0.12
    0.11
    762
    +0.10
    0.14
    477
    +0.10
    0.04
    Negative Logits
     frustration
    -0.76
     frustrations
    -0.74
     disadvantage
    -0.74
    uyomi
    -0.73
    efficiency
    -0.72
     worries
    -0.71
     tedious
    -0.71
    igslist
    -0.69
     dissatisf
    -0.68
     impat
    -0.68
    POSITIVE LOGITS
     existed
    1.13
    ewitness
    1.08
     indeed
    1.04
     actually
    0.92
     murdered
    0.89
    actually
    0.87
     originated
    0.87
     EVER
    0.85
     alive
    0.84
     faked
    0.83
    Activations Density 4.554%

    No Known Activations