Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 24576
    5. 8-RES_FS24576-JB
    6. 23292
    Prev
    Next
    INDEX
    Explanations

    the word "but" and its variations, highlighting contrast or exceptions in statements

    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
    66
    +0.11
    0.5%
    483
    +0.10
    0.5%
    740
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    66
    +0.11
    0.03
    233
    +0.10
    0.03
    740
    +0.09
    0.03
    Negative Logits
    ento
    -0.88
    oire
    -0.85
    ORPG
    -0.67
    oleon
    -0.67
    utonium
    -0.66
    swick
    -0.65
    axter
    -0.65
    OLOGY
    -0.65
    anwhile
    -0.64
    icipated
    -0.62
    POSITIVE LOGITS
     nonetheless
    1.25
     nevertheless
    1.19
     unmist
    0.96
    chery
    0.94
     effective
    0.94
     undeniably
    0.93
     undeniable
    0.91
     manageable
    0.90
    tons
    0.89
     still
    0.88
    Activations Density 0.066%

    No Known Activations