Kedis ku xidhan hagaha · Dhexdhexaad Heerka

Kediska Soo-saarista Modelka iyo Weerarada Xatooyada

Fahmo sida weerarradu ay ugu daydaan moodooyinka sanduuqa madow iyada oo loo marayo weydiimo, waxa API-ga soo saarayaa ay muujinayaan, iyo sida bixiyayaasha u ogaan karaan ama u joojin karaan soo saarista.

Wadooyinka haga ee la xidhiidhaQaabka Soo Saaridda Iyo Weerarada Xatooyada
Su'aal 1 ee 8

Waa maxay fikradda udub-dhexaadka u ah weerarka soo saarista moodeelka?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Waa maxay fikradda udub-dhexaadka u ah weerarka soo saarista moodeelka?
    • Weydiinta qaabka iyo tababbarka beddelka si uu ugu daydo wax soo saarkiisa
    • jabsiga server-ka si aad si toos ah u koobiyayso faylasha miisaanka
    • Laaluushka shaqaalaha si uu u daadiyo xogta tababarka
    • Si xad dhaaf ah u raritaanka API-ga ilaa ay ka burburto oo ay tuurto xusuusta
  2. Maxay tahay sababta soo celinta ixtimaalka buuxa ay uga dhigayso soo saarista ka sahlan soo celinta summada-1-sare oo keliya?
    • Waxay siraysaa jawaabta, iyadoo hoos u dhigaysa weerarka
    • Way soo daadataa macluumaad dheeraad ah su'aal kasta, markaa su'aalo yar ayaa loo baahan yahay
    • Waxay qarinaysaa gebi ahaanba xadka go'aanka
    • Waxay kordhisaa kharashka server-ka si ay weeraryahanadu u quustaan
  3. Waa kuwee farsamada isdifaaca oo si toos ah u yaraynaysa macluumaadka la daadiyay weydiimihii kasta?
    • Ku-daridda GPU-yo badan serferka
    • Kordhinta tirada cabbirka moodeelka
    • Soo celinta wax soo saarka wareegsan ama ugu sarreeya-1-keliya
    • Daabacaadda xogta tababarka si furan
  4. What did the 2024 production-language-model extraction paper report about GPT-3.5 Turbo?
    • It reconstructs every GPT-3.5 parameter exactly for under $20
    • It obtains the model’s training dataset directly
    • It extracts a complete model checkpoint from the API
    • It recovers the embedding-projection layer; the paper estimates under $2,000 in queries for GPT-3.5 Turbo
  5. Muxuu moodalka beddelka ah ee la xaday khatar ugu yahay dakhliga lumay?
    • Waxay awood u siinaysaa weerarrada sanduuqa cad ee khadka tooska ah sida samaynta tusaalayaal iska soo horjeeda
    • Waxay wanaajisaa saxnaanta moodeelkii asalka ahaa
    • Waxay si toos ah u tirtirtaa qaabka asalka ah
    • Waxaa loo baahan yahay in dib loo tababaro asalka
  6. In ActiveThief, what does the combined DFAL + K-center query strategy aim to balance?
    • Boundary-proximate informative samples and reduced redundancy among selected samples
    • Exact recovery of all secret model weights and a guaranteed substitute
    • Random sampling only and avoiding all substitute-model predictions
    • Higher API throughput and removal of the query budget
  7. What can API rate limits do against a model-extraction attempt?
    • Prove that no substitute model can be trained
    • Remove all information from each permitted prediction
    • Make large-scale querying slower or more expensive
    • Make model outputs impossible to compare
  8. What does the DynaMarks paper propose for a model-extraction defense?
    • It adds a logo to every generated answer
    • It encrypts all model weights at rest
    • It guarantees ownership of every copied production model
    • It can embed a signature in a surrogate model by dynamically changing API outputs using secret parameters