© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    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. 3800
    Prev
    Next
    INDEX
    Explanations

    online forum replies that include specific functions or actions like replying and logging in

    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
    50
    +0.10
    0.5%
    326
    +0.09
    0.4%
    679
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    23
    +0.10
    0.22
    352
    +0.09
    0.20
    154
    +0.09
    0.19
    Negative Logits
    apping
    -0.73
    orsi
    -0.72
    iders
    -0.70
    ardi
    -0.65
    cape
    -0.65
    ookie
    -0.64
    coni
    -0.63
    女
    -0.62
    iris
    -0.62
    milo
    -0.62
    POSITIVE LOGITS
     Delete
    0.93
    gments
    0.81
    gements
    0.72
    gment
    0.70
    gement
    0.67
    Reply
    0.66
    ged
    0.61
    wra
    0.60
    PRESS
    0.60
    ative
    0.59
    Activations Density 5.238%

    No Known Activations