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

    Neuronpedia

    APISteerSAE EvalsBlog/PodcastNEWSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlog/PodcastGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 768
    5. 8-RES_FS768-JB
    6. 162
    Prev
    Next
    INDEX
    Explanations

    instances of the word "even" being used in a sentence to emphasize a particular point

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    718
    +0.12
    0.6%
    163
    +0.11
    0.5%
    253
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    718
    +0.12
    0.10
    133
    +0.11
    0.08
    253
    +0.10
    0.08
    Negative Logits
    rend
    -0.99
    aim
    -0.88
    asus
    -0.83
    utra
    -0.79
    othy
    -0.74
    ugal
    -0.73
    cel
    -0.73
    ubi
    -0.73
    HCR
    -0.71
    uments
    -0.68
    POSITIVE LOGITS
     though
    1.07
     tho
    0.98
     remotely
    0.95
    handed
    0.94
    handedly
    0.84
     worse
    0.79
     curs
    0.73
     if
    0.69
     stranger
    0.68
     moderately
    0.64
    Activations Density 2.398%

    No Known Activations