© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 8913
    Prev
    Next
    INDEX
    Explanations

    the conjunction "but" used to introduce contrasting ideas or exceptions

    oai_token-act-pair · gpt-4o-mini
    Triggered 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
    Neuron Alignment
    Index
    Value
    % of L₁
    760
    +0.11
    0.5%
    102
    +0.10
    0.5%
    175
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    66
    +0.11
    0.03
    740
    +0.10
    0.03
    233
    +0.10
    0.03
    Negative Logits
    ento
    -0.73
    eatures
    -0.69
    oire
    -0.66
    anamo
    -0.65
    rea
    -0.65
    bia
    -0.64
    OLOGY
    -0.62
    vana
    -0.62
    arta
    -0.61
    anwhile
    -0.61
    POSITIVE LOGITS
     persistent
    0.91
     effective
    0.89
     unsuccessful
    0.83
    chery
    0.83
     durable
    0.83
     nonetheless
    0.82
     unmist
    0.82
     lethal
    0.80
     measurable
    0.79
     accurate
    0.79
    Activations Density 0.154%

    No Known Activations