© 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 EvalsExportsinterp-engineNEWAPI 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. 16747
    Prev
    Next
    INDEX
    Explanations

    words related to communication or verbal interaction

    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
    532
    +0.11
    0.5%
    191
    +0.10
    0.5%
    235
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    532
    +0.11
    0.05
    380
    +0.10
    0.03
    282
    +0.10
    0.04
    Negative Logits
    abase
    -1.02
    atcher
    -0.79
    unity
    -0.73
    reviewed
    -0.72
    osponsors
    -0.72
    vre
    -0.67
     MFT
    -0.66
    rift
    -0.66
    itton
    -0.66
    agate
    -0.65
    POSITIVE LOGITS
     aloud
    1.38
     plaint
    1.22
     uttered
    1.12
     loudly
    1.11
     words
    1.10
     cryptic
    1.09
     loud
    1.05
     phrases
    1.04
     greeting
    1.00
     softly
    0.99
    Activations Density 0.411%

    No Known Activations