© 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. 23875
    Prev
    Next
    INDEX
    Explanations

    references to misunderstandings or misconceptions

    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
    720
    +0.12
    0.5%
    591
    +0.11
    0.5%
    448
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    448
    +0.12
    0.03
    140
    +0.11
    0.03
    736
    +0.11
    0.03
    Negative Logits
    acho
    -0.75
    agra
    -0.70
    idal
    -0.68
    itaire
    -0.68
    lite
    -0.68
    mone
    -0.67
    hedral
    -0.65
    inder
    -0.65
    ternity
    -0.63
    lov
    -0.63
    POSITIVE LOGITS
     perceptions
    0.94
     impression
    0.82
     assumptions
    0.80
     overest
    0.76
     inaccur
    0.75
     inaccurate
    0.75
     misunderstanding
    0.75
     underestimate
    0.74
     misconceptions
    0.73
     stereotypes
    0.72
    Activations Density 0.085%

    No Known Activations