© 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 49152
    9. 8-RES_FS49152-JB
    11. 37318
    Prev
    Next
    INDEX
    Explanations

    modal verbs expressing possibility or uncertainty

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    524
    +0.10
    0.5%
    221
    +0.10
    0.5%
    195
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    551
    +0.10
    0.05
    605
    +0.10
    0.04
    278
    +0.10
    0.03
    Negative Logits
    ament
    -0.79
    efeated
    -0.75
    rency
    -0.69
    cies
    -0.69
    fighting
    -0.69
    ventures
    -0.68
    treated
    -0.67
     Fighter
    -0.65
    athing
    -0.65
    bread
    -0.65
    POSITIVE LOGITS
     misunder
    1.09
     someday
    0.86
    hap
    0.85
     imply
    0.84
     offend
    0.83
     mitigate
    0.83
     subconscious
    0.82
     incent
    0.81
     confuse
    0.80
     discourage
    0.80
    Activations Density 0.063%

    No Known Activations