© 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
    2. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    3. GPT2-Small
    4. Residual Post - 32k
    5. 10-RES_POST_32K-OAI
    6. 29583
    Prev
    Next
    INDEX
    Explanations

    references to magical concepts and supernatural elements

    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
    288
    +0.13
    0.6%
    642
    +0.10
    0.4%
    738
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.13
    0.04
    401
    +0.10
    0.03
    404
    +0.09
    0.02
    Negative Logits
    Filename
    -0.91
    ership
    -0.74
    aton
    -0.71
    bis
    -0.70
    lr
    -0.68
     Racer
    -0.67
    Stage
    -0.67
    agonists
    -0.66
    etsk
    -0.66
    greSQL
    -0.66
    POSITIVE LOGITS
     realism
    1.03
     wand
    0.94
     lantern
    0.90
     potions
    0.90
     potion
    0.89
     tricks
    0.88
     mushrooms
    0.86
     carpet
    0.84
    binding
    0.81
     spells
    0.80
    Activations Density 0.202%

    No Known Activations