© 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 3072
    9. 8-RES_FS3072-JB
    11. 3020
    Prev
    Next
    INDEX
    Explanations

    phrases related to investigative journalism or controversial 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
    459
    +0.16
    0.8%
    503
    +0.11
    0.5%
    501
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    452
    +0.16
    0.07
    450
    +0.11
    0.07
    598
    +0.10
    0.05
    Negative Logits
     constitu
    -0.70
    ËĪ
    -0.67
    iatus
    -0.66
    NetMessage
    -0.65
     ILCS
    -0.64
    ocaust
    -0.64
     âĵĺ
    -0.64
     Flavoring
    -0.63
    ATIONAL
    -0.62
     dstg
    -0.62
    POSITIVE LOGITS
    sequently
    1.16
     ensuing
    1.05
     subsequently
    0.96
    sequent
    0.91
     latter
    0.87
     subsequent
    0.86
     thereafter
    0.85
     resultant
    0.84
     later
    0.82
     afterward
    0.80
    Activations Density 3.652%

    No Known Activations