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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Under Peer Review · Attention SAE Research Paper
    5. GPT2-Small
    7. Attention Out
    9. 2-ATT-KK
    11. 22731
    Prev
    Next
    INDEX
    Explanations

    references to the metal "brass" and related terms such as "knuckles" and "plates."

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

    references to brass objects or terms associated with military or authoritative contexts

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 2-att-kk
    Configuration
    ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    float32
    Hook Name
    blocks.2.attn.hook_z
    Hook Layer
    2
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Head Attr Weights
    0:0.10
    1:0.05
    2:0.15
    3:0.07
    4:0.05
    5:0.08
    6:0.03
    7:0.04
    8:0.06
    9:0.23
    10:0.06
    11:0.03
    Negative Logits
     Discrimination
    -1.33
    owan
    -1.29
     Pradesh
    -1.27
    Dispatch
    -1.23
     Fiona
    -1.21
     Pengu
    -1.21
    BuyableInstoreAndOnline
    -1.21
     Pix
    -1.20
     Citiz
    -1.20
     adolesc
    -1.19
    POSITIVE LOGITS
     robes
    1.40
     helmets
    1.36
    ebus
    1.22
     pads
    1.17
     hood
    1.14
    ero
    1.14
     watches
    1.12
     helm
    1.12
     Bullets
    1.10
    atron
    1.10
    Activations Density 0.002%

    No Known Activations