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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 768
    5. 8-RES_FS768-JB
    6. 424
    Prev
    Next
    INDEX
    Explanations

    phrases related to overcoming challenges and perseverance

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    71
    +0.10
    0.4%
    581
    +0.09
    0.4%
    214
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    581
    +0.10
    0.09
    163
    +0.09
    0.08
    735
    +0.09
    0.08
    Negative Logits
    ahime
    -0.78
    eka
    -0.68
    omas
    -0.68
    instein
    -0.67
    OGR
    -0.65
    qus
    -0.64
    hua
    -0.64
    edia
    -0.64
    ello
    -0.62
    rers
    -0.60
    POSITIVE LOGITS
     except
    1.51
     imaginable
    1.38
    except
    1.17
     conceivable
    0.99
     whatsoever
    0.96
    Except
    0.81
     alike
    0.81
    together
    0.79
     besides
    0.78
     irrespective
    0.78
    Activations Density 4.624%

    No Known Activations