© 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 EvalsExportsinterp-engineNEWAPI 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. 27242
    Prev
    Next
    INDEX
    Explanations

    sequences of colons and associated patterns, likely indicating a focus on formatting or coding syntax

    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
    326
    +0.13
    0.6%
    107
    +0.10
    0.5%
    160
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    326
    +0.13
    0.02
    160
    +0.10
    0.03
    701
    +0.10
    0.01
    Negative Logits
     coh
    -0.80
     haunting
    -0.73
     tatt
    -0.73
     aven
    -0.73
     afore
    -0.72
     stalking
    -0.72
     tort
    -0.71
     confir
    -0.69
    quished
    -0.68
     indo
    -0.68
    POSITIVE LOGITS
    operator
    1.34
    iterator
    1.17
    NL
    1.15
    Base
    1.10
    basic
    1.05
    Union
    1.04
    GEN
    1.04
    temp
    1.02
    Internal
    1.02
    before
    1.02
    Activations Density 0.092%

    No Known Activations