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 Local
    5. 6-RES_SCL-AJT
    6. 45413
    Prev
    Next
    INDEX
    Explanations

    descriptions and discussions related to legal or investigative actions

    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
    +0.26
    1.3%
    321
    +0.09
    0.4%
    430
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    618
    +0.26
    0.19
    640
    +0.09
    0.32
    430
    +0.09
    0.25
    Negative Logits
     achieves
    -0.67
     dominates
    -0.66
     intrinsically
    -0.64
     disproportionately
    -0.64
     consumes
    -0.63
     endif
    -0.61
     hits
    -0.61
     consume
    -0.60
     multipl
    -0.60
     survives
    -0.59
    POSITIVE LOGITS
     asking
    0.73
     requesting
    0.72
    ername
    0.70
    iscons
    0.67
     inquire
    0.65
    perty
    0.65
    mbuds
    0.65
    rompt
    0.64
    Monitor
    0.64
    alion
    0.64
    Activations Density 22.859%

    No Known Activations