© 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. 4775
    Prev
    Next
    INDEX
    Explanations

    words and phrases related to harm and consequences in interpersonal contexts

    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
    1:0.06
    2:0.13
    3:0.05
    4:0.03
    5:0.05
    6:0.21
    7:0.04
    8:0.04
    9:0.21
    10:0.04
    11:0.03
    Negative Logits
     CY
    -3.78
    izoph
    -3.43
     Spect
    -3.23
    gran
    -3.14
     precinct
    -3.12
    ================================
    -3.11
    ambers
    -3.10
     clutch
    -3.10
     Nanto
    -3.04
     Sanford
    -3.04
    POSITIVE LOGITS
     hurt
    5.63
     Hurt
    5.40
     harmed
    5.21
     hurts
    5.16
     hurting
    4.98
     Harm
    4.93
     harm
    4.66
     harms
    4.65
     harming
    4.57
     injure
    4.29
    Activations Density 0.003%

    No Known Activations