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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Apollo Research · Taylor · Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning
    3. GPT2-Small
    4. Residual Similar CE E2E Recon
    5. 6-RES_SCEFR-AJT
    6. 576
    Prev
    Next
    INDEX
    Explanations

    phrases indicating permission or conditionality, such as "provided" or "granting."

    oai_token-act-pair · gpt-3.5-turbo
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    neuronpedia/gpt2-small__res_scefr-ajt/6-res_scefr-ajt
    Prompts (Dashboard)
    12,288 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    46,080
    Data Type
    torch.float32
    Hook Point
    blocks.6.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    apollo-research/Skylion007-openwebtext-tokenizer-gpt2
    Hook Point Layer
    6
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    642
    +0.17
    0.6%
    739
    +0.16
    0.5%
    59
    +0.14
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    217
    +0.17
    0.04
    310
    +0.16
    0.04
    59
    +0.14
    0.04
    Negative Logits
    focus
    -1.13
    psc
    -1.08
    pred
    -1.07
    bold
    -0.97
    step
    -0.96
    ipel
    -0.96
    mer
    -0.95
    kos
    -0.94
    tra
    -0.94
    âĸ¬
    -0.93
    POSITIVE LOGITS
    DragonMagazine
    1.14
    utical
    1.08
     supplied
    0.93
    emonium
    0.93
     refunds
    0.93
     warranties
    0.91
     generously
    0.91
     herein
    0.90
     Ez
    0.90
     receipt
    0.85
    Activations Density 0.316%

    No Known Activations