© 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. 10256
    Prev
    Next
    INDEX
    Explanations

    mentions of numbers or quantities, especially tens of thousands

    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
    71
    +0.12
    0.5%
    393
    +0.10
    0.4%
    169
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    71
    +0.12
    0.03
    344
    +0.10
    0.04
    689
    +0.10
    0.03
    Negative Logits
    avior
    -0.80
    dict
    -0.70
    icut
    -0.66
    NET
    -0.66
    Creat
    -0.65
    Fram
    -0.65
    Land
    -0.64
    ahime
    -0.64
    kaya
    -0.64
    obal
    -0.63
    POSITIVE LOGITS
     thousands
    1.60
     miles
    1.20
     millions
    1.19
     thousand
    1.16
     Thousands
    1.09
     pounds
    1.08
     kilometres
    1.08
     kilometers
    1.07
     acres
    1.04
     thous
    1.03
    Activations Density 0.108%

    No Known Activations