© 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. 796
    Prev
    Next
    INDEX
    Explanations

    references to legal actions or investigations

    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
    283
    +0.10
    0.5%
    24
    +0.09
    0.5%
    458
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    283
    +0.10
    0.04
    526
    +0.09
    -0.03
    570
    +0.08
    0.00
    Negative Logits
     reminds
    -0.68
     wonders
    -0.58
     nutshell
    -0.55
     partName
    -0.55
     tells
    -0.55
    iple
    -0.55
     HERE
    -0.54
     comes
    -0.53
     loves
    -0.52
     brings
    -0.52
    POSITIVE LOGITS
     ".
    0.65
    etheless
    0.64
     respectively
    0.61
    )).
    0.60
    ]).
    0.59
     undet
    0.59
     ineligible
    0.58
     theirs
    0.58
    itored
    0.57
     amounted
    0.57
    Activations Density 2.901%

    No Known Activations