© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. Under Peer Review · Attention SAE Research Paper
    5. GPT2-Small
    7. Attention Out
    9. 3-ATT-KK
    11. 14418
    Prev
    Next
    INDEX
    Explanations

    quotes and statements related to government policies and public sectors

    oai_token-act-pair · gpt-3.5-turbo
    Triggered by @bot

    expressions related to political positions and actions

    oai_token-act-pair · gpt-4o-mini
    Triggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 3-att-kk
    Configuration
    ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    float32
    Hook Name
    blocks.3.attn.hook_z
    Hook Layer
    3
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Head Attr Weights
    0:0.07
    1:0.12
    2:0.07
    3:0.02
    4:0.11
    5:0.10
    6:0.09
    7:0.07
    8:0.08
    9:0.06
    10:0.07
    11:0.07
    Negative Logits
     Craigslist
    -1.47
    ???
    -1.38
     harbor
    -1.36
     los
    -1.34
     goddamn
    -1.34
     someday
    -1.33
     Neighbor
    -1.30
     $$
    -1.29
     Brooklyn
    -1.28
     Hawai
    -1.28
    POSITIVE LOGITS
     criticised
    2.05
     recognise
    2.02
     Nato
    1.89
     centre
    1.84
     campaigners
    1.81
     realise
    1.79
     centres
    1.78
     recogn
    1.78
     scept
    1.77
     apologised
    1.76
    Activations Density 0.610%

    No Known Activations