© 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. 4665
    Prev
    Next
    INDEX
    Explanations

    the word "how" with a high level of activation

    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
    +0.12
    0.6%
    333
    +0.10
    0.5%
    699
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    55
    +0.12
    0.02
    192
    +0.10
    0.02
    699
    +0.10
    0.02
    Negative Logits
    urrence
    -0.71
     Flavoring
    -0.66
    76561
    -0.66
    aceae
    -0.64
    holder
    -0.63
    anmar
    -0.63
    ãĤ¨ãĥ«
    -0.58
    UME
    -0.58
    Afee
    -0.58
    abol
    -0.57
    POSITIVE LOGITS
     much
    0.95
     messed
    0.84
     fatig
    0.83
     stren
    0.82
     diligent
    0.78
     badly
    0.78
     insignificant
    0.77
     MUCH
    0.77
    ls
    0.77
     crappy
    0.75
    Activations Density 0.036%

    No Known Activations