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 24576
    5. 8-RES_FS24576-JB
    6. 23364
    Prev
    Next
    INDEX
    Explanations

    commands or calls to action related to halting or ceasing activities

    oai_token-act-pair · gpt-4o-miniTriggered 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
    693
    +0.11
    0.5%
    517
    +0.10
    0.4%
    759
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    665
    +0.11
    0.04
    669
    +0.10
    0.04
    766
    +0.09
    0.03
    Negative Logits
    uin
    -0.71
    amen
    -0.69
     liner
    -0.65
    izable
    -0.64
     recovers
    -0.64
    reset
    -0.64
    albeit
    -0.64
    Luckily
    -0.62
    istically
    -0.61
    save
    -0.61
    POSITIVE LOGITS
     wasting
    0.97
     discriminatory
    0.92
     abusive
    0.92
     whining
    0.87
     wasteful
    0.86
     worrying
    0.83
     Discrimination
    0.81
     harassing
    0.81
     excessive
    0.80
     hypocrisy
    0.80
    Activations Density 0.217%

    No Known Activations