© 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 1536
    9. 8-RES_FS1536-JB
    11. 11
    Prev
    Next
    INDEX
    Explanations

    negative words or phrases, particularly related to criticism or problems

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    36
    +0.33
    1.7%
    393
    +0.17
    0.9%
    263
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    36
    +0.33
    0.40
    250
    +0.17
    0.16
    159
    +0.13
    0.16
    Negative Logits
    anism
    -0.81
    Edit
    -0.73
    evidence
    -0.72
    coins
    -0.69
    agents
    -0.69
    Contents
    -0.68
    Examples
    -0.67
    blocks
    -0.66
    agree
    -0.66
    grounds
    -0.66
    POSITIVE LOGITS
     lot
    1.45
     plethora
    1.35
     bunch
    1.31
     handful
    1.22
     slew
    1.18
     multitude
    1.16
     whopping
    1.12
     few
    1.11
     couple
    1.09
     sizable
    1.08
    Activations Density 1.696%

    No Known Activations