© 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 24576
    9. 8-RES_FS24576-JB
    11. 1710
    Prev
    Next
    INDEX
    Explanations

    sentences describing a process of change or growth

    oai_token-act-pair · gpt-3.5-turboTriggered 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
    393
    +0.13
    0.6%
    36
    +0.13
    0.6%
    512
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    36
    +0.13
    0.08
    512
    +0.13
    0.05
    263
    +0.11
    0.04
    Negative Logits
    agree
    -0.75
     fears
    -0.65
     beliefs
    -0.64
     convictions
    -0.63
     desires
    -0.63
     allegedly
    -0.62
     allegations
    -0.62
    atures
    -0.61
     evidently
    -0.61
     notions
    -0.60
    POSITIVE LOGITS
     boon
    1.06
     fruitful
    1.06
     catalyst
    0.84
     beneficial
    0.84
     someday
    0.83
     wiser
    0.82
     decisive
    0.80
     useful
    0.80
     deterrent
    0.78
     seamless
    0.78
    Activations Density 0.182%

    No Known Activations