Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 3072
    5. 8-RES_FS3072-JB
    6. 1918
    Prev
    Next
    INDEX
    Explanations

    verbs related to a competition or comparison, particularly emphasizing the act of winning or outperforming

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs3072-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    3,072
    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
    181
    +0.10
    0.4%
    750
    +0.09
    0.4%
    178
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    750
    +0.10
    0.05
    3
    +0.09
    0.05
    123
    +0.09
    0.05
    Negative Logits
     Relief
    -0.80
    pmwiki
    -0.77
    itory
    -0.68
    ":"/
    -0.68
    atto
    -0.63
     coh
    -0.61
    orrow
    -0.61
    raine
    -0.60
     eman
    -0.60
     fallout
    -0.59
    POSITIVE LOGITS
     expectations
    0.88
     rivals
    0.76
    ingly
    0.76
    ealous
    0.75
    beat
    0.72
     rival
    0.72
    xual
    0.72
     beaten
    0.71
     opponents
    0.70
     odds
    0.69
    Activations Density 0.907%

    No Known Activations