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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 3072
    5. 8-RES_FS3072-JB
    6. 2562
    Prev
    Next
    INDEX
    Explanations

    mentions of software, European security concerns, justice, discriminatory practices, implementation of systems and policies, working conditions, abuse, gender identity and expression, delays, water, seas, theories, professional conduct, consumption, aspirations, services, teachings, principles, protocols, open concepts, pensions, and legal arrangements

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs3072-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    3,072
    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
    79
    +0.12
    0.6%
    762
    +0.11
    0.5%
    504
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.12
    0.16
    762
    +0.11
    0.15
    225
    +0.10
    0.13
    Negative Logits
    pload
    -0.66
    aed
    -0.66
    atus
    -0.61
    otin
    -0.61
    cus
    -0.60
    amaz
    -0.60
    antha
    -0.60
    kernel
    -0.59
    imum
    -0.59
    omaly
    -0.59
    POSITIVE LOGITS
     alike
    1.91
     respectively
    1.40
     thereof
    0.84
     sectors
    0.82
     industries
    0.80
     depending
    0.80
     versa
    0.77
    depending
    0.73
    SPONSORED
    0.73
     combo
    0.65
    Activations Density 2.190%

    No Known Activations