Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    3. GPT2-Small
    4. Attention Out - 128k
    5. 0-ATT_128K-OAI
    6. 13552
    Prev
    Next
    INDEX
    Explanations

    expressions of respect and disrespect

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    jbloom/GPT2-Small-OAI-v5-128k-attn-out-SAEs/v5_128k_layer_0
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    131,072
    Data Type
    torch.float32
    Hook Name
    blocks.0.hook_attn_out
    Hook Layer
    0
    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
    624
    +0.20
    1.3%
    183
    +0.18
    1.2%
    62
    +0.16
    1.0%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    393
    +0.20
    0.10
    284
    +0.18
    0.09
    257
    +0.16
    0.09
    Negative Logits
    Virginia
    -0.89
    Dust
    -0.81
    ilk
    -0.80
    peria
    -0.78
    outhern
    -0.77
    aints
    -0.77
    isconsin
    -0.77
    Torrent
    -0.74
    Hyper
    -0.73
    MD
    -0.72
    POSITIVE LOGITS
     corrid
    0.85
    VERTISEMENT
    0.85
     oun
    0.82
     concess
    0.80
     sacrific
    0.80
    SHIP
    0.77
    æĸ¹
    0.75
     metic
    0.75
     respect
    0.74
     thous
    0.74
    Activations Density 1.831%

    No Known Activations