HyDE Document Hypothétique
HyDE dafay gëna yombal seetlu bi ci laaj modelu làkk bi mu xalaat ab këyitu tontu bu baaxul, ba noppi seetee ak këyit biñu samp ci barabu laaj bu ñor bi.
Résumé
It bridges the gap between short questions and the longer passages you actually want to find.
Plongeur bu xóot
HyDE (Hypothetical Document Embeddings), bi Gao ak ay naataango yi xalaat ci 2022, dafay xeex jafe-jafe ci seetlu bu dëgër: ab laaj bu gàtt ak ab tontu bu am solo dañuy faral di dëkk ci gox yu wuute ci espace embedding. Reset bi amna ñatti jéego. Bi njëkk mooy nga laaj LLM bi topp tegtal yi (lu melni InstructGPT) mu defar benn këyit buy tontu laaj bi, doonte dafa am ay leeral yuñ sosal wala yu jaarul yoon. Ñaareel ba mooy nga dugal këyitu xalaat boobu ak enkodeer buy wuutale te kenn duko saytu (lu melni Contriever). Ñatteel ba mooy jëfandikoo embedding bi ngir gis passage dëgg ci seetug dëkkandoo bi gëna jege. Encoder bi dafay melni compresseur buy ñàkk, di segg defar LLM yi ci noonu muy tëye siñaal semantik bi war. Li yéeme mooy HyDE dafay dox zero-shot, soxlawul ay done yuñ etikete, te dafay méngoo wala di raw retrievers yuñ defar bu baax ci làkk yi ak ci liggéey yi.
Gis-gis xarala
Li am solo mooy ni jéego bi ñuy dugal dafay denoiser bu bari bruit. Doonte këyit biñ defar mën na am ay njuumte ci dëgg, encoder bu dëgër bi daf koy méngale ak ay pàcc yu am solo ndax dañu bokk motif topique ak semantik, ci noonu lañuy raxas ay mbir yu amul benn werante ci biir bottleneck bu vecteur bu am dayo fixe. HyDE dafay soppi bërëb bi ci tàggat ab encoder laajte ngir jàppale xam-xam bi LLM di defar ak ab embedder bu nekk ci biti te kenn duko saytu.
njeextalu pexe
Gaawaay ak yaatuwaay
Liggéeyukaay yi ci làkk yi mën nañu gëna gaaw te duñu yàq deggoo gi.
Dugg ak yegg
Dafay yaatal jëfandikoo gi ci làkk yi ak ci anam yi ñuy jokkoo.
dogal yu gëna leer
Ekip yi mën nañu gëna yàgg ci àtte ci jamono ji otomatisation di liggéey ci baamtu.
Ëlëgu HyDE Document Hypothétique
HyDE ab bloku tabax la ci pipeline RAG yu xarañ yi, ñu koy faral di boole ak reranking ak defar ay laaj yu bari. Xaarandi ay variant yuy defar ay këyitu hypothétique yu bari te moyenne seen embeddings ngir robustness, jëfandikoo adaptive buy trigger HyDE rek sudee laaj bu ñor bi baaxul, ak lëkkaloo bu gëna dëgër ak LLMs local yu gëna xéewale ngir wàññi latency ak njëg. Lu xeetu defar yi di gëna yokk, kalite këyitu xalaat yi - ak seetlu - dafa wara wéy di yokk.
Doxal ci àdduna dëgg
Seetug zero-shot ci domen bu bees bu amul benn done buñ etiketee ci wàllu laaj-passage
Seetug làkk yu bari, defar ab tontu bu nuruwul ci làkk wiñ la bindal balaa ngay dugal
Yokkatal fàttaliku RAG ci yaatal laaji jëfandikukat yu gàtt yi ci pseudo-document yu am solo
Gëstu ak seetlu yoon fu laaj yu gàtt yi wara méngoo ak pàcc yu dëgër, yu bari jargon
Risk yi ak balustrade yi
Lépp lu jaarul yoon mën na dugg ci rapoor yi, jàppale ci liggéey bi, wala ci njariñu gëstu bi.
Sensibilite bu gaaw mën na jur njariñ yu wuute ci laajte yu noonu mel.
Done yu am solo mën nañu feeñ sudee seytu jëfandikoo gi néew doole.
Roadmap ngir samp gi
Mandargal formaa génne gi, melokaan bi, ak standard kalite yi laata ngay dugal ko.
Tontu yu am solo ak balluwaay yu wóor saa yu dëggu bi di am solo.
Fexeel am barabu xool nit ñi ngir am njariñ yu am solo.
Toppal anami gacce yi ak di faral di tàggataat ay laaj wala def-liggéey.
Weyal di banneexu
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the HyDE Hypothetical Document Embeddings quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Gis bi ci topp
FastText Subword
Laaj yi ñuy faral di laaj
What is HyDE Hypothetical Document Embeddings?
HyDE dafay gëna yombal seetlu bi ci laaj modelu làkk bi mu xalaat ab këyitu tontu bu baaxul, ba noppi seetee ak këyit biñu samp ci barabu laaj bu ñor bi. Dafay dindi bërëb bi nekk ci diggante laaj yu gàtt yi ak pàcc yu gëna gudd yi nga bëgga gis.
Lan la HyDE di defar balaa muy amal retrieval?
HyDE dafay laaj LLM mu bind ab këyitu tontu buñu xalaat, buñu ko jëfandikoo ngir seet ay pàcc yu dëggu.
Lu tax amul benn njariñ sudee këyit bi ñuy xalaat dafa am njuumte ci dëgg?
Embedding bu am dayo fixe dafay nuru bottleneck buy jàpp relevance aktual bi ci di sànni ay detay yu baaxul.
Ban xeetu modelu embedding lañu jëfandikoo ci HyDE bi njëkk ngir encoder këyitu hypothétique bi?
HyDE dafay boole generatëru LLM bi ak enkodeer buy wuutale buñu dul saytu bu melni Contriever ngir samp këyitu xalaat bi.
Ban jafe-jafe bu mag ci retrieval dense la HyDE nara saafara?
Soo soppi laajte bi ci mbind mu niru ak këyit, HyDE dafay toxal vecteur seetlu bi mu gëna jege xeetu paas yi mu wara méngoo.