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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & 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. 1610
    Prev
    Next
    INDEX
    Explanations

    companies and related concepts such as corporate entities and startups

    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
    75
    +0.11
    0.5%
    269
    +0.11
    0.5%
    167
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    75
    +0.11
    0.07
    146
    +0.11
    0.06
    269
    +0.11
    0.05
    Negative Logits
    lyn
    -0.75
    loe
    -0.71
    itives
    -0.71
    bian
    -0.70
    ipeg
    -0.68
    cue
    -0.68
    oubt
    -0.67
    alam
    -0.65
    izoph
    -0.65
    iggins
    -0.65
    POSITIVE LOGITS
     executives
    1.10
     headquartered
    1.07
     CEOs
    0.96
     subsidiary
    0.94
    orpor
    0.90
     specializing
    0.87
     owned
    0.84
     subsidiaries
    0.83
     CEO
    0.83
     founders
    0.83
    Activations Density 0.603%

    No Known Activations