© 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. 8-RES_POST_32K-OAI
    6. 6708
    Prev
    Next
    INDEX
    Explanations

    references to subscription details and timeframes

    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
    496
    +0.10
    0.5%
    162
    +0.09
    0.4%
    315
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    238
    +0.10
    0.02
    39
    +0.09
    0.01
    651
    +0.09
    0.02
    Negative Logits
     imperson
    -0.83
     spitting
    -0.64
     Skydragon
    -0.63
     halluc
    -0.62
    osate
    -0.61
     skiing
    -0.59
     melting
    -0.58
    NULL
    -0.58
    cffffcc
    -0.58
    ega
    -0.58
    POSITIVE LOGITS
     Extend
    0.78
    Extra
    0.70
     Includes
    0.70
    Beg
    0.67
    plus
    0.67
     tier
    0.66
    starting
    0.65
    start
    0.65
    renheit
    0.65
    "}],"
    0.64
    Activations Density 0.070%

    No Known Activations