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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Apollo Research · Taylor · Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning
    5. GPT2-Small
    7. Residual Similar CE E2E Recon
    9. 6-RES_SCEFR-AJT
    11. 13675
    Prev
    Next
    INDEX
    Explanations

    mentions of laws, policies, and regulations

    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
    •Layer 6 UMAP region: e2e-recon cluster above bottome line touching local cluster - e2e-recon

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    679
    +0.25
    1.0%
    640
    +0.15
    0.6%
    440
    +0.13
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    679
    +0.25
    0.23
    6
    +0.15
    0.21
    640
    +0.13
    0.20
    Negative Logits
     beasts
    -0.96
     extinct
    -0.88
     beast
    -0.86
     naughty
    -0.83
     invincible
    -0.82
     silent
    -0.81
     docks
    -0.80
     voy
    -0.79
     backwards
    -0.79
     fronts
    -0.79
    POSITIVE LOGITS
     Similarly
    1.12
     Consequently
    1.11
     However
    1.10
     Afterwards
    1.09
     Later
    1.08
     Previously
    1.08
     Nonetheless
    1.06
     Likewise
    1.05
     Moreover
    1.05
     Nevertheless
    1.04
    Activations Density 2.652%

    No Known Activations