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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Under Peer Review · Attention SAE Research Paper
    5. GPT2-Small
    7. Attention Out
    9. 3-ATT-KK
    11. 12655
    Prev
    Next
    INDEX
    Explanations

    scientific studies and related information involving universities or researchers

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot

    mentions of universities and their associated personnel or entities

    oai_token-act-pair · gpt-4o-miniTriggered 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
    Not in Any Lists

    No Comments

    Head Attr Weights
    0:0.02
    1:0.03
    2:0.08
    3:0.02
    4:0.03
    5:0.07
    6:0.28
    7:0.13
    8:0.07
    9:0.05
    10:0.12
    11:0.05
    Negative Logits
    ilib
    -1.17
    stro
    -1.13
    uce
    -1.08
     paradise
    -1.07
    party
    -1.06
    enfranch
    -1.06
    iband
    -1.03
    etimes
    -1.02
    erie
    -1.02
    polit
    -1.02
    POSITIVE LOGITS
     clerk
    1.18
     Instructor
    1.18
     Mellon
    1.14
     Clerk
    1.13
     Edwin
    1.13
     examiner
    1.12
     Savannah
    1.05
     Dept
    1.05
     transcripts
    1.05
     Transcript
    1.04
    Activations Density 0.051%

    No Known Activations