© 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 24576
    9. 8-RES_FS24576-JB
    11. 1706
    Prev
    Next
    INDEX
    Explanations

    words related to extreme outcomes or intensifiers

    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
    335
    +0.11
    0.5%
    343
    +0.10
    0.5%
    101
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    335
    +0.11
    0.02
    158
    +0.10
    0.01
    109
    +0.10
    0.01
    Negative Logits
    bars
    -0.82
    ades
    -0.79
    ker
    -0.79
    rooms
    -0.76
    then
    -0.75
    ohn
    -0.74
    crow
    -0.72
    oles
    -0.72
     Mub
    -0.71
    etimes
    -0.70
    POSITIVE LOGITS
     gratification
    1.01
     prize
    0.97
     goal
    0.96
     embodiment
    0.95
     culmination
    0.91
     obstacle
    0.87
     irony
    0.85
     pinnacle
    0.84
     showdown
    0.84
     solution
    0.83
    Activations Density 0.021%

    No Known Activations