© 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. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 17757
    Prev
    Next
    INDEX
    Explanations

    phrases related to incorporating new information or evidence

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    411
    +0.09
    0.4%
    12
    +0.09
    0.4%
    68
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    761
    +0.09
    0.04
    469
    +0.09
    0.03
    526
    +0.09
    0.00
    Negative Logits
    Ĥİ
    -0.72
     Correspond
    -0.67
    guard
    -0.63
     tending
    -0.61
    Uncommon
    -0.60
     Machina
    -0.59
    edom
    -0.58
    journal
    -0.58
     FIGHT
    -0.58
    bard
    -0.57
    POSITIVE LOGITS
     into
    1.59
     Into
    1.41
     INTO
    1.40
    into
    1.33
     onto
    1.17
    tion
    0.83
     overboard
    0.81
     seamlessly
    0.71
    interstitial
    0.71
    rafted
    0.69
    Activations Density 0.186%

    No Known Activations