© 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 49152
    9. 8-RES_FS49152-JB
    11. 26794
    Prev
    Next
    INDEX
    Explanations

    positive descriptions of individuals, particularly in terms of personality and ability

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    762
    +0.11
    0.5%
    392
    +0.10
    0.4%
    438
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    762
    +0.11
    0.04
    45
    +0.10
    0.02
    56
    +0.10
    0.03
    Negative Logits
     Versus
    -0.86
    ebin
    -0.86
    soDeliveryDate
    -0.81
    ackets
    -0.78
    imester
    -0.73
    IRE
    -0.71
    ARA
    -0.70
    Ĥİ
    -0.70
     Recall
    -0.69
    atari
    -0.68
    POSITIVE LOGITS
     loved
    1.03
     admired
    1.02
    loving
    1.00
     valued
    0.99
     loving
    0.93
     caring
    0.92
     wonderful
    0.92
     compassionate
    0.91
     loves
    0.90
     respected
    0.89
    Activations Density 0.182%

    No Known Activations