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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports 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. 5514
    Prev
    Next
    INDEX
    Explanations

    conditional phrases involving "or" and "and"

    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
    276
    +0.11
    0.5%
    272
    +0.11
    0.5%
    69
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    393
    +0.11
    0.11
    484
    +0.11
    0.08
    225
    +0.09
    0.09
    Negative Logits
    éĹĺ
    -0.72
    aturated
    -0.65
    gian
    -0.62
    illon
    -0.59
    eteria
    -0.58
    unci
    -0.58
    essor
    -0.57
    doms
    -0.57
    imates
    -0.57
    Ö¼
    -0.57
    POSITIVE LOGITS
     purposes
    1.12
     amusement
    1.09
     sake
    1.02
     enjoyment
    0.89
     souven
    0.89
     entertainment
    0.88
     gratification
    0.84
     convenience
    0.84
     novelty
    0.81
     enlightenment
    0.81
    Activations Density 1.136%

    No Known Activations