© 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. 10-RES_POST_32K-OAI
    11. 30746
    Prev
    Next
    INDEX
    Explanations

    the conjunctions and connective phrases within sentences

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.10.hook_resid_post
    Hook Layer
    10
    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
    547
    +0.12
    0.5%
    576
    +0.11
    0.5%
    86
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    484
    +0.12
    0.08
    86
    +0.11
    0.08
    393
    +0.10
    0.10
    Negative Logits
    ================================
    -0.78
     =================================
    -0.76
    ********************************
    -0.76
    âĢ¦âĢ¦âĢ¦âĢ¦
    -0.74
     ðŁĻĤ
    -0.72
    ~~~~
    -0.71
     (=
    -0.71
     ;)
    -0.70
     :-)
    -0.70
    ~~~~~~~~
    -0.70
    POSITIVE LOGITS
     urged
    1.45
     vowed
    1.42
     warned
    1.24
     said
    1.17
     likened
    1.13
     hinted
    1.13
     cautioned
    1.12
     criticized
    1.12
     pledged
    1.12
     blamed
    1.10
    Activations Density 0.924%

    No Known Activations