Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 3072
    5. 8-RES_FS3072-JB
    6. 2591
    Prev
    Next
    INDEX
    Explanations

    details about events and actions happening in certain places

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs3072-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    3,072
    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
    484
    +0.12
    0.5%
    734
    +0.11
    0.5%
    636
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    636
    +0.12
    0.07
    362
    +0.11
    0.08
    717
    +0.10
    0.06
    Negative Logits
     rooft
    -0.71
    é¾įå
    -0.66
     props
    -0.65
     surfaces
    -0.65
     deserts
    -0.63
     cats
    -0.62
    é£
    -0.61
     backdrop
    -0.61
     uploaded
    -0.60
     coefficients
    -0.59
    POSITIVE LOGITS
     countdown
    1.01
     fray
    0.87
     hostilities
    0.85
     dialogue
    0.84
     anew
    0.81
     journey
    0.80
    Phase
    0.76
     conversation
    0.76
    uberty
    0.75
     discussion
    0.73
    Activations Density 1.392%

    No Known Activations