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. 13525
    Prev
    Next
    INDEX
    Explanations

    instances of the word "perform" in various contexts related to actions or activities taking place

    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
    7
    +0.23
    1.7%
    167
    +0.16
    1.2%
    527
    +0.15
    1.1%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    455
    +0.23
    0.02
    253
    +0.16
    0.01
    70
    +0.15
    0.01
    Negative Logits
    EStream
    -0.99
    PsyNetMessage
    -0.92
    76561
    -0.86
    ablishment
    -0.80
     msec
    -0.75
    bleacher
    -0.74
     ATK
    -0.71
    addafi
    -0.71
    ï¸ı
    -0.68
    dden
    -0.67
    POSITIVE LOGITS
    Ü
    0.93
    etics
    0.86
    port
    0.86
    inance
    0.83
    ible
    0.81
    eway
    0.81
    alities
    0.80
    anooga
    0.80
    inances
    0.80
    ibble
    0.79
    Activations Density 0.028%

    No Known Activations