© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 1536
    9. 8-RES_FS1536-JB
    11. 102
    Prev
    Next
    INDEX
    Explanations

    phrases related to causes and effects

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    608
    +0.14
    0.7%
    481
    +0.13
    0.6%
    650
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    608
    +0.14
    0.15
    703
    +0.13
    0.08
    362
    +0.12
    0.11
    Negative Logits
    ura
    -0.84
    ires
    -0.82
    ishi
    -0.79
    imensional
    -0.79
    iac
    -0.79
    heit
    -0.78
    enth
    -0.78
    lest
    -0.76
    jah
    -0.75
    okes
    -0.75
    POSITIVE LOGITS
     sheer
    1.26
     lack
    1.11
     advancements
    1.10
     inexper
    1.08
     inaction
    1.07
     fears
    1.04
     unforeseen
    1.03
     exhaustion
    0.97
     insufficient
    0.95
     factors
    0.94
    Activations Density 2.325%

    No Known Activations