© 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. 45603
    Prev
    Next
    INDEX
    Explanations

    the presence of words with specific vowel or consonant patterns

    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
    373
    +0.12
    0.6%
    116
    +0.12
    0.5%
    526
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    738
    +0.12
    0.01
    373
    +0.12
    -0.00
    116
    +0.10
    0.01
    Negative Logits
    ortunately
    -0.78
    ĪĴ
    -0.78
     Bened
    -0.77
    «ĺ
    -0.73
    pta
    -0.69
     Pwr
    -0.69
    Ĥª
    -0.66
     Vie
    -0.66
    Ļ½
    -0.65
    quartered
    -0.65
    POSITIVE LOGITS
    kids
    1.14
    Kids
    1.11
    arro
    0.91
    kid
    0.89
    bang
    0.86
    ARDS
    0.86
    illas
    0.83
    hawk
    0.75
    busters
    0.74
    apons
    0.73
    Activations Density 0.013%

    No Known Activations