Kediska Hallucinations AI

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Why does next-token prediction alone fail to verify a factual claim in a generated answer?
   • Every model is connected to only unreliable websites
   • All users ask grammatically incorrect questions
   • A plausible continuation is not the same as evidence that the statement is true
   • A fluent answer automatically proves the fact
2. Waa maxay sababta ay dhalanteedka u yihiin khatar gaar ah?
   • Had iyo jeer waxay ku qoran yihiin naxwaha jaban si isticmaalayaashu ay iska indhatiraan
   • Waxay kaliya ka muuqdaan fariimaha khaldan
   • Waxay sababaan in moodelku burburo
   • Waxaa loo dhiibaa si faseexa oo kalsooni leh, sidaas darteed qaladaadku waa sahlan tahay in la seego
3. Jiilka la kordhiyey ee dib u soo celinta (RAG) miyay gabi ahaanba baabi'inaysaa dhalanteedka?
   • Haa, RAG waxay dammaanad qaadaysaa jawaabaha dhabta ah
   • Maya, moodooyinka wali way is khilaafi karaan ama iska indhatirin karaan qoraalka la soo saaray ee buuqa ama iska hor imanaya
   • Haa, laakiin kaliya maalmaha fasaxa ah
   • Maya, sababtoo ah RAG waligeed ma bixiso wax dukumeenti ah
4. Waa kuwee tusaale dhab ah oo dhalanteed ah?
   • Qaabka oo daliishanaya kiis maxkamadeed oo aan jirin
   • Qaabka oo diiday inuu ka jawaabo su'aal
   • Qaabka ayaa weydiinaya su'aal caddaynaysa
   • Qaabka oo soo celinaya jawaab madhan
5. Which workflow can help check a generated factual answer?
   • Trust every citation because it looks formal
   • Treat a low sampling temperature as proof of correctness
   • Hide the evidence so the answer is easier to read
   • Retrieve relevant material and verify whether its statements and citations support the claims
6. How should a claim that a new model has eliminated hallucinations be assessed?
   • Accept it solely because the announcement sounds confident
   • Assume one successful example proves every application
   • Use defined tasks, evidence, and measurement conditions rather than assume universal reliability
   • Judge only whether the model produces polished prose
7. A generated answer cites a real page, but the page does not support the associated factual statement. What remains unverified?
   • The support for that particular claim
   • Whether a URL can be opened
   • Whether the model used fluent sentences
   • Whether the citation contains text
8. A supplied document says delivery takes five business days, but its generated summary says two. Which check reveals the error?
   • Counting how many sentences the summary contains
   • Comparing the summary's quantity with the source document
   • Checking only whether the prose sounds confident
   • Assuming summaries may replace any source number