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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & 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. 6477
    Prev
    Next
    INDEX
    Explanations

    references to the "Star Wars" franchise and related concepts

    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
    554
    +0.25
    1.9%
    283
    +0.23
    1.7%
    208
    +0.16
    1.3%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    326
    +0.25
    0.01
    530
    +0.23
    0.01
    67
    +0.16
    0.01
    Negative Logits
    istries
    -1.12
    bestos
    -1.00
    uras
    -1.00
    paio
    -0.99
    clud
    -0.94
    izoph
    -0.94
    natureconservancy
    -0.94
    ntil
    -0.91
    manuel
    -0.91
    inventoryQuantity
    -0.90
    POSITIVE LOGITS
     Selling
    0.85
     Frequ
    0.81
     Twice
    0.80
     Scenes
    0.80
     Duration
    0.77
     Massacre
    0.76
     Talks
    0.75
     Beyond
    0.75
     Worldwide
    0.73
     Combined
    0.72
    Activations Density 0.023%

    No Known Activations