© 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 12288
    9. 8-RES_FS12288-JB
    11. 9881
    Prev
    Next
    INDEX
    Explanations

    phrases related to investigations, audits, reports, and findings

    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
    481
    +0.20
    0.9%
    100
    +0.10
    0.4%
    291
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    481
    +0.20
    0.10
    150
    +0.10
    0.04
    291
    +0.09
    0.06
    Negative Logits
    iku
    -0.98
    iversary
    -0.90
     ç¥ŀ
    -0.81
    swer
    -0.80
    izons
    -0.80
    enment
    -0.79
     rejoice
    -0.79
     tomorrow
    -0.79
    Tomorrow
    -0.77
    iour
    -0.77
    POSITIVE LOGITS
     improperly
    1.25
     misconduct
    1.24
     unlawfully
    1.23
     violations
    1.23
     improper
    1.18
     negligence
    1.18
     defendants
    1.17
     negligent
    1.14
     mishand
    1.13
     inappropriately
    1.11
    Activations Density 1.361%

    No Known Activations