© 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
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 12288
    5. 8-RES_FS12288-JB
    6. 9621
    Prev
    Next
    INDEX
    Explanations

    phrases related to legal matters and criminal activities, particularly those involving arrests, charges, and controversies

    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
    314
    +0.13
    0.6%
    694
    +0.13
    0.6%
    280
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.13
    0.09
    251
    +0.13
    0.05
    694
    +0.10
    0.04
    Negative Logits
    soDeliveryDate
    -0.89
    ellen
    -0.80
     wisely
    -0.76
    isSpecialOrderable
    -0.74
    izons
    -0.72
     Studios
    -0.65
    Stud
    -0.64
     Cannes
    -0.63
     Tee
    -0.63
     Zuckerberg
    -0.63
    POSITIVE LOGITS
     violations
    1.18
     poisoning
    1.17
     theft
    1.15
     contamination
    1.14
     disturbance
    1.11
     misuse
    1.10
    angering
    1.10
     disturbances
    1.07
     injuries
    1.06
     wrongdoing
    1.05
    Activations Density 0.742%

    No Known Activations