© 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 12288
    9. 8-RES_FS12288-JB
    11. 1392
    Prev
    Next
    INDEX
    Explanations

    phrases related to walking down a street or hallway

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    459
    +0.11
    0.5%
    708
    +0.11
    0.5%
    123
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    113
    +0.11
    0.04
    123
    +0.11
    0.04
    469
    +0.10
    0.03
    Negative Logits
    ²¾
    -0.87
    manship
    -0.85
    rss
    -0.73
    FORE
    -0.73
    achev
    -0.72
    thood
    -0.70
    ortunate
    -0.69
    ilic
    -0.69
    tight
    -0.68
    tle
    -0.68
    POSITIVE LOGITS
     halls
    1.16
     streets
    0.98
     corridors
    0.95
     rooft
    0.86
     hall
    0.84
     sidewalks
    0.83
     hallway
    0.81
     doors
    0.79
     gates
    0.78
     woods
    0.77
    Activations Density 0.312%

    No Known Activations