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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & 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 Local
    5. 6-RES_SCL-AJT
    6. 34136
    Prev
    Next
    INDEX
    Explanations

    references to religious figures, especially the Prophet Muhammad, and locations such as mosques

    oai_token-act-pair · gpt-3.5-turbo
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    neuronpedia/gpt2-small__res_scl-ajt/6-res_scl-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
    447
    +0.20
    1.1%
    288
    +0.13
    0.7%
    247
    +0.09
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    362
    +0.20
    0.13
    473
    +0.13
    0.03
    755
    +0.09
    0.05
    Negative Logits
     rink
    -0.56
    ucha
    -0.56
    ensing
    -0.56
    ateral
    -0.54
    egal
    -0.53
    elsius
    -0.53
    tten
    -0.52
    YC
    -0.52
    monton
    -0.52
    rown
    -0.52
    POSITIVE LOGITS
     Muhammad
    0.85
     Mohammed
    0.75
     Elijah
    0.74
     Jeremiah
    0.72
     Mohammad
    0.71
     Amos
    0.71
     Isaiah
    0.71
    esses
    0.70
     Prophet
    0.68
     Ezra
    0.67
    Activations Density 10.671%

    No Known Activations