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

    Neuronpedia

    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. 1460
    Prev
    Next
    INDEX
    Explanations

    past-tense verbs that indicate an action or achievement accomplished

    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
    55
    +0.14
    0.6%
    418
    +0.11
    0.5%
    215
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    55
    +0.14
    0.05
    418
    +0.11
    0.04
    215
    +0.11
    0.04
    Negative Logits
     Notting
    -0.69
     lash
    -0.64
    phabet
    -0.63
    culosis
    -0.63
     compose
    -0.63
    erity
    -0.60
    ously
    -0.60
     flair
    -0.59
     Patrol
    -0.59
     lapse
    -0.58
    POSITIVE LOGITS
     bothered
    0.96
     traced
    0.95
     forgiven
    0.91
     construed
    0.90
     undone
    0.89
     accommod
    0.87
     trusted
    0.86
    amed
    0.82
     reconcil
    0.81
     blamed
    0.81
    Activations Density 0.076%

    No Known Activations