© 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. 15925
    Prev
    Next
    INDEX
    Explanations

    phrases describing a significant impact or emphasis on a particular subject

    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
    447
    +0.11
    0.5%
    84
    +0.10
    0.4%
    714
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    714
    +0.11
    0.02
    106
    +0.10
    0.02
    61
    +0.09
    0.02
    Negative Logits
    LER
    -0.84
    anth
    -0.78
    mons
    -0.73
    ricanes
    -0.70
    atures
    -0.67
    nings
    -0.67
    yip
    -0.66
    ilk
    -0.66
     Springs
    -0.65
    yle
    -0.65
    POSITIVE LOGITS
     unchanged
    1.05
     overlooked
    1.03
     reliant
    0.96
     unaffected
    0.93
     unnoticed
    0.93
     unexpl
    0.90
     ignored
    0.89
     forgotten
    0.87
     absent
    0.86
     ineffective
    0.85
    Activations Density 0.025%

    No Known Activations