© 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. 5040
    Prev
    Next
    INDEX
    Explanations

    mentions of different levels of severity or intensity in events or situations

    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
    103
    +0.11
    0.5%
    492
    +0.10
    0.5%
    288
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    232
    +0.11
    0.02
    103
    +0.10
    0.01
    392
    +0.10
    0.01
    Negative Logits
    CHR
    -0.80
    senal
    -0.72
     Penal
    -0.69
     Skydragon
    -0.65
     Schwarzenegger
    -0.61
     ANGEL
    -0.60
    OPER
    -0.60
    ²¾
    -0.59
     fear
    -0.58
     Trin
    -0.58
    POSITIVE LOGITS
    ield
    1.24
    irmation
    1.06
    iculty
    1.05
    erences
    1.05
    irming
    1.04
    irms
    1.03
    onso
    1.03
    irmed
    1.03
    ront
    1.01
    erent
    1.00
    Activations Density 0.014%

    No Known Activations