© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI 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. 41644
    Prev
    Next
    INDEX
    Explanations

    negative or cautionary statements regarding political actions and their consequences

    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
    457
    +0.10
    0.5%
    291
    +0.09
    0.4%
    567
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    393
    +0.10
    0.03
    170
    +0.09
    0.02
    546
    +0.09
    0.02
    Negative Logits
     Previously
    -0.76
    Ö¼
    -0.75
    querque
    -0.71
    iliated
    -0.70
    igslist
    -0.68
    ĵ
    -0.68
    Later
    -0.68
    arag
    -0.68
    ãĤ¶
    -0.68
    onds
    -0.67
    POSITIVE LOGITS
     ALWAYS
    1.01
     whichever
    0.94
     regardless
    0.87
     whatever
    0.85
     always
    0.84
     invariably
    0.81
     nevertheless
    0.79
     irrespective
    0.78
     whatsoever
    0.76
    whatever
    0.75
    Activations Density 0.122%

    No Known Activations