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 3072
    5. 8-RES_FS3072-JB
    6. 1221
    Prev
    Next
    INDEX
    Explanations

    phrases related to political news and events

    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
    242
    +0.13
    0.6%
    261
    +0.10
    0.5%
    627
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    242
    +0.13
    0.07
    627
    +0.10
    0.06
    164
    +0.10
    0.07
    Negative Logits
     smugglers
    -0.77
     manuals
    -0.73
    Reviewer
    -0.72
    tics
    -0.71
     ransom
    -0.69
     condoms
    -0.69
     lookout
    -0.68
     terrorists
    -0.68
     compilation
    -0.67
     rhy
    -0.66
    POSITIVE LOGITS
     Pennsylvania
    1.05
     Idaho
    1.03
    Virginia
    1.01
    Calif
    1.01
    Florida
    1.01
    Portland
    1.01
    Michigan
    1.00
     Ohio
    0.99
     Minnesota
    0.99
    California
    0.99
    Activations Density 0.717%

    No Known Activations