© 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 12288
    9. 8-RES_FS12288-JB
    11. 9584
    Prev
    Next
    INDEX
    Explanations

    mentions of the word "crystal"

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    261
    +0.11
    0.5%
    651
    +0.10
    0.5%
    525
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    326
    +0.11
    0.03
    288
    +0.10
    0.03
    525
    +0.10
    0.03
    Negative Logits
    gets
    -0.78
    orship
    -0.78
    esp
    -0.73
    ername
    -0.72
    dit
    -0.72
    erness
    -0.72
    nir
    -0.72
    HAHAHAHA
    -0.67
    ahah
    -0.67
    alach
    -0.67
    POSITIVE LOGITS
    stals
    0.90
     latt
    0.85
     Gems
    0.84
     Palace
    0.82
    toe
    0.81
     meth
    0.81
     MX
    0.80
     quartz
    0.77
    stal
    0.77
    ined
    0.75
    Activations Density 0.136%

    No Known Activations