Encodeurs croisés vs Bi-encodeurs
Ñaari anam la modele neuronal yi di méngale mbind: bi-encoder yi dañuy boole piyees bu nekk ngir seetlu bu gaaw, waaye cross-encoder yi dañuy boole mbind yépp ngir gëna dëggu.
Résumé
The choice shapes the speed-versus-precision tradeoff in every modern search and retrieval system.
Plongeur bu xóot
Ñari architecture yi tontu 'ban la ñaari mbind di lëkkaloo ?', waaye wuute nañu ci kañ la mbind yi daje. Bi-encoder dafay doxal frase bu nekk ci transformateur bi moom boppam, muy defar benn vecteur fixe ci bind bu nekk; nuru mooy produit point bu yomb wala cosinus ci digganté vecteur yi. Ndax mën nañu njëkka xayma vecteur yi ba noppi denc leen, bi-encoder yi mën nañu dem ba yegg ci ay milioŋ ciy këyit ak base de done vecteur yu am doole. Encoder cross-encoder dafay boole ñaari mbind yi ([CLS] laaj [SEP] këyit) ba noppi boole leen ci model bi, bàyyi token bu nekk mu topp beneen token bu nekk balaa muy génne benn poñ bu am solo. Bii xel yépp dafay jàpp jaxasoo yu sew yi bi-encoder bi ñàkk, kon cross-encoder yi ñoo gëna dëggu waaye mënu ñu precompute dara te dañu wara dox benn yoon ci peer bi.
Gis-gis xarala
Seen wuutu mooy yaatuwaayu bàyyi xel. Ci bi-encoder, self-attention du musa jàkkarloo ak ñaari duggal yi, kon dugal këyitu dokimaa mënul laaj te mën nañu ko jëfandikoowaat. Ci biir cross-encoder, bàyyi xel dafay jaar ci toppalante biñ boole, moo tax poñ yi dañu aju ci laaj bi. Njëg yi dañuy méngoo ak loolu: rang N këyitu soxla N paas transformatër bu mat ngir encodeur cross ak N méngale vecteur yu yomb ngir bi-encodeur ginaaw benn encode laaj.
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 encodeur cross vs Bi-encodeur
Modèle bi gëna am solo mooy hybrid recuperation-then-rank: bi-encoder dafay jël ay téemeeri kàndidaa ci ay milioŋ, ginaaw ga cross-encoder dafay raññewaat resultaa yi gëna kawe. Modèle interaction yu mujj yu melni ColBERT dañu xaaj wuute gi ci denc vecteur token bu nekk, distilaasioŋ bi dafay gëna tàggat bi-encoder yu kompact ngir toppandoo àtte encoder cross. Xaarandi rerankers yu gëna xéewale ak boole bu gëna dëgër ci ñaari etap yi ci gasoduk defar yu gëna yokk.
Doxal ci àdduna dëgg
Base de done vecteur dafay jëfandikoo bi-encoder ngir am 200 pàcc yi gëna am solo ci ay milioŋ ciy këyit ci ay milisegond
Benn cross-encoder ranker dafay raññewaat 200 kàndidaa yi balaa ñu leen di jox benn chatbot RAG, loolu mooy gëna yombal njariñu tontu
Transformers-Transformers yi dañuy yónnee ay encodeur yuñ tàggat bu njëkk (ngir seetlu semantik) ak ay encodeur yu jàkkarloo (ngir rang bu bees ak joxe poñ STS)
Gis ñaari laaj ci forum Q&A dafay jëfandikoo encoder cross ngir méngale ñaar ci limu gàtt
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 Cross-Encoders vs Bi-Encoders 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
BERT ak model encoder
Laaj yi ñuy faral di laaj
What is Cross-Encoders vs Bi-Encoders?
Ñaari anam la modele neuronal yi di méngale mbind: bi-encoder yi dañuy boole piyees bu nekk ngir seetlu bu gaaw, waaye cross-encoder yi dañuy boole mbind yépp ngir gëna dëggu. Tanneef bi mooy tëral njariñu gaawaay ak njub ci bépp sistemu seetlu ak wut bu bees.
Lan mooy wuutale architecture bi am ci digganté encodeur croisé ak encodeur bi-encodeur?
Cross-encoder yi dañuy boole ñaari dugal yi ba noppi bàyyi xel ci toppalante bi yépp; bi-encoder yi dañuy kode mbind mu nekk ci ay vecteur yu wuute.
Lan moo waral bi-encoder yi di mëna dem ba yegg ci ay milioŋ ciy këyit ci anam wu jaar yoon?
Ndax dokimaa bu nekk dafay enkode boppam, vecteur bi mën nañu ko jëfandikoowaat ci bépp laaj te mën nañu ko indexe bu njëkk.
Ci gasoduk seetug liggéey buñ miin, naka lañuy boole ñaari architecture yi?
Modèlu recuperation-puis-rérank buñ miin dafay jëfandikoo bi-encoder bu gaaw ngir fàttaliku lu yaatu ak cross-encoder bu jaar yoon ngir raññewaat limu gàtt yi.
Lu tax cross-encoder mënul njëkka xayma misaali këyit?
Ginaaw poñ yi ñu ngi bawoo ci dokimaa-laaj biñ boole, mingi aju ci laaj-te dañu ko wara xaymaat ci peer bu nekk.
Lan la bi-encoder di génne ci benn mbind buñ dugal?
Bi-encoder dafay boole mbind mu nekk ci benn vecteur buñ samp; suko defee ñu xayma nuru gi ci digganté vecteur yi.