© 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 12288
    9. 8-RES_FS12288-JB
    11. 1407
    Prev
    Next
    INDEX
    Explanations

    political and governmental terms

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    402
    +0.11
    0.5%
    602
    +0.10
    0.5%
    440
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    720
    +0.11
    0.05
    602
    +0.10
    0.04
    546
    +0.10
    0.04
    Negative Logits
    gae
    -0.67
    inventoryQuantity
    -0.64
    DX
    -0.61
    */(
    -0.61
    iatus
    -0.61
    lies
    -0.59
    Latest
    -0.58
    aturdays
    -0.55
    Def
    -0.53
    ãĤ£
    -0.53
    POSITIVE LOGITS
    would
    1.55
     would
    1.46
     wouldn
    1.42
     Would
    1.39
     Wouldn
    1.20
    Would
    1.12
    'd
    1.03
    could
    0.98
     could
    0.88
     risked
    0.82
    Activations Density 0.918%

    No Known Activations