© 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. 1615
    Prev
    Next
    INDEX
    Explanations

    phrases related to a specific area in a vehicle, hinting at it being the cockpit

    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
    288
    +0.17
    0.8%
    418
    +0.12
    0.5%
    659
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.17
    0.03
    652
    +0.12
    0.02
    678
    +0.11
    0.02
    Negative Logits
     Berks
    -0.79
    District
    -0.72
    reddits
    -0.71
    utical
    -0.70
    inos
    -0.70
    berman
    -0.70
    teness
    -0.69
    govtrack
    -0.69
    女
    -0.68
    venants
    -0.66
    POSITIVE LOGITS
     compartment
    1.16
     cabin
    1.01
     hatch
    1.01
     cockpit
    0.98
     cavity
    0.92
     depress
    0.88
     windshield
    0.87
     liner
    0.84
     attendants
    0.83
     ceiling
    0.83
    Activations Density 0.047%

    No Known Activations