© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    5. GPT2-Small
    7. Residual Post - 32k
    9. 8-RES_POST_32K-OAI
    11. 6767
    Prev
    Next
    INDEX
    Explanations

    phrases related to conditional and proportional relationships, often in the context of distribution or valuation

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.8.hook_resid_post
    Hook Layer
    8
    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
    566
    +0.12
    0.5%
    421
    +0.12
    0.5%
    702
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    665
    +0.12
    0.06
    469
    +0.12
    0.04
    195
    +0.10
    0.02
    Negative Logits
    flies
    -0.83
    tek
    -0.75
    alter
    -0.71
    oho
    -0.70
    thia
    -0.70
    stadt
    -0.69
    ãĥīãĥ©
    -0.69
    essa
    -0.68
    void
    -0.67
    uve
    -0.66
    POSITIVE LOGITS
     downwards
    1.03
     differently
    0.98
     favour
    0.96
    orously
    0.93
     favorably
    0.93
     accordingly
    0.93
     appropriately
    0.92
     upwards
    0.91
     upward
    0.90
     downward
    0.86
    Activations Density 0.527%

    No Known Activations