© 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 24576
    9. 8-RES_FS24576-JB
    11. 7165
    Prev
    Next
    INDEX
    Explanations

    pronouns referring to people or entities in a negative context

    oai_token-act-pair · gpt-3.5-turboTriggered 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
    139
    +0.14
    0.6%
    571
    +0.12
    0.5%
    156
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    139
    +0.14
    0.06
    571
    +0.12
    0.05
    566
    +0.11
    0.04
    Negative Logits
     Eleven
    -0.78
     Greatest
    -0.71
     Around
    -0.68
     Dayton
    -0.66
    icion
    -0.63
    Republic
    -0.62
     Hague
    -0.62
     Worse
    -0.61
    UGE
    -0.61
     Tyr
    -0.61
    POSITIVE LOGITS
     lacked
    0.95
     lacks
    0.92
     didn
    0.90
     knew
    0.86
     dared
    0.86
     got
    0.82
     feared
    0.82
     survived
    0.82
    've
    0.81
     didnt
    0.80
    Activations Density 0.163%

    No Known Activations