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. 13556
    Prev
    Next
    INDEX
    Explanations

    references to the time of day, specifically focusing on "evening."

    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
    336
    +0.28
    2.2%
    178
    +0.21
    1.7%
    624
    +0.15
    1.2%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    336
    +0.28
    0.02
    152
    +0.21
    0.03
    504
    +0.15
    0.02
    Negative Logits
    ãĥ¼ãĥĨãĤ£
    -0.80
    PsyNetMessage
    -0.79
    76561
    -0.77
    0000000
    -0.76
    ++)
    -0.73
    ipation
    -0.72
    raltar
    -0.72
    kHz
    -0.71
    EXP
    -0.70
    ufficient
    -0.70
    POSITIVE LOGITS
    ctica
    0.92
     Lumpur
    0.91
    agher
    0.86
    gradation
    0.81
    Ħ¢
    0.80
    ŃĶ
    0.79
    izons
    0.78
    agraph
    0.77
    phabet
    0.76
    mediately
    0.74
    Activations Density 0.054%

    No Known Activations