© 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. 9-ATT-KK
    11. 16155
    Prev
    Next
    INDEX
    Explanations

    contextual references regarding Detroit's current status and reputation, often comparing perceptions and realities

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 9-att-kk
    Configuration
    ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L9_Hcat_z_lr1.20e-03_l11.20e+00_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.9.attn.hook_z
    Hook Layer
    9
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Head Attr Weights
    0:0.07
    1:0.02
    2:0.03
    3:0.03
    4:0.08
    5:0.03
    6:0.22
    7:0.05
    8:0.07
    9:0.29
    10:0.02
    11:0.04
    Negative Logits
     Bates
    -4.50
     Cruz
    -3.84
     ski
    -3.74
     Sultan
    -3.63
    Cruz
    -3.63
     Banana
    -3.59
     Slater
    -3.57
     nun
    -3.51
     Spectre
    -3.38
     Ashe
    -3.38
    POSITIVE LOGITS
    Detroit
    8.95
     Detroit
    8.67
     DET
    5.39
     Flint
    5.03
     Pistons
    4.97
     Lions
    4.86
     Michigan
    4.64
    Michigan
    4.56
     Stafford
    4.42
     Lansing
    4.13
    Activations Density 0.009%

    No Known Activations