© 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. 10-ATT-KK
    11. 16499
    Prev
    Next
    INDEX
    Explanations

    phrases indicating negative intentions or malicious behavior

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-att-kk
    Configuration
    ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L10_Hcat_z_lr1.20e-03_l11.30e+00_ds24576_bs4096_dc1.00e-05_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.10.attn.hook_z
    Hook Layer
    10
    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
    1:0.14
    2:0.03
    3:0.05
    4:0.08
    5:0.03
    6:0.10
    7:0.15
    8:0.03
    9:0.04
    10:0.06
    11:0.15
    Negative Logits
    opian
    -2.98
     Tycoon
    -2.52
    Surv
    -2.42
     Tale
    -2.41
    uko
    -2.29
     monarch
    -2.27
    assic
    -2.26
     Deng
    -2.25
     lockdown
    -2.23
    aq
    -2.21
    POSITIVE LOGITS
     intent
    3.56
    intent
    3.25
    vous
    3.02
     Intent
    3.00
     sincerity
    2.90
     intentions
    2.82
     malice
    2.75
     PUR
    2.74
     Purpose
    2.73
     intending
    2.70
    Activations Density 0.002%

    No Known Activations