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. 6472
    Prev
    Next
    INDEX
    Explanations

    types of animated creatures and themes of struggle or conflict

    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
    99
    +0.24
    1.9%
    151
    +0.22
    1.7%
    288
    +0.20
    1.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    99
    +0.24
    0.02
    326
    +0.22
    0.01
    288
    +0.20
    0.02
    Negative Logits
    ukong
    -0.93
    ocular
    -0.91
    ographers
    -0.87
    perture
    -0.84
    aird
    -0.80
    ovie
    -0.80
    pta
    -0.80
    opsis
    -0.78
    ographer
    -0.77
    kson
    -0.76
    POSITIVE LOGITS
    etheless
    0.91
     disadvant
    0.82
     Awakening
    0.79
     denomin
    0.76
    tarian
    0.76
    tarians
    0.73
    manship
    0.71
     mascul
    0.70
    iciary
    0.68
    hips
    0.68
    Activations Density 0.023%

    No Known Activations