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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 1536
    9. 8-RES_FS1536-JB
    11. 83
    Prev
    Next
    INDEX
    Explanations

    keywords related to construction, renovation, and urban planning

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    325
    +0.15
    0.7%
    487
    +0.11
    0.5%
    216
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    325
    +0.15
    0.10
    288
    +0.11
    0.13
    42
    +0.10
    0.09
    Negative Logits
    sweet
    -0.79
    uel
    -0.75
    zee
    -0.72
     Fren
    -0.72
    andra
    -0.72
    dos
    -0.70
    ican
    -0.69
    wild
    -0.68
    vana
    -0.68
     Cele
    -0.66
    POSITIVE LOGITS
     bridges
    1.00
     constructed
    1.00
     scaff
    0.96
     erected
    0.96
    built
    0.92
    itect
    0.90
    Constructed
    0.89
     builders
    0.89
     construction
    0.86
     dams
    0.86
    Activations Density 2.246%

    No Known Activations