Kunna Kai Mai Kyau-Tuning
Daidaita wasa da kai yana inganta ƙirar ta hanyar yin gogayya da ko koyi daga abubuwan da ya gabata, samar da siginar horo.
Dubawa
It matters because it can push performance beyond the supervised data using little or no extra human labeling.
Zurfafa nutsewa
Wasan kai yana da tushe mai zurfi a cikin wasan AI: AlphaGo Zero da AlphaZero sun isa wasan ɗan adam kawai ta hanyar kunna miliyoyin wasanni akan kansu, ba tare da rikodin wasan ɗan adam ba. Irin wannan ruhu yanzu yana bayyana a cikin ƙirar harshe mai kyau. A cikin SPIN (Self-Play fine-tuNing), samfurin na yanzu yana haifar da martani ga faɗakarwa, kuma horo yana tura ƙirar don bambanta amsoshin da aka samar daga ainihin waɗanda aka rubuta na ɗan adam, ɗaukar kanta a matsayin duka mai kunnawa da abokin gaba. Bayan maimaita maimaitawa 'abokin gaba' (madaidaicin wurin bincike na baya) yana ƙara ƙarfi, don haka ƙirar dole ne ta ci gaba da inganta, sannu a hankali yana rufe rata tare da rarraba manufa. Babban abin roko shine ingancin bayanai: ana iya matse tsayayyen saitin bayanai don ƙarin riba ba tare da tattara sabbin zanga-zangar ɗan adam ko abubuwan da ake so ba.
Fahimtar Fasaha
SPIN yana tsara kyakkyawan daidaitawa azaman wasan 'yan wasa biyu tare da asarar salon DPO: ƙirar an horar da ita don ba da mafi girman yuwuwar ga martanin tunani na ɗan adam fiye da nasa wanda ya ƙirƙiri kansa daga abubuwan da suka gabata. Saboda wurin binciken da ya gabata yana ba da abubuwan da ba su da kyau, wahalar yana daidaitawa ta atomatik yayin da ƙirar ta inganta. A cikin tsarin wasan-wasa, ana haɗa wasan-kai tare da bincike (misali, MCTS) da cibiyar sadarwa mai ƙima, samar da tsari mara iyaka na abokan adawar ci gaba ba tare da bayanan waje ba.
Dabarun Tasiri
Shawarwari masu haske
Yana taimaka muku keɓance bayyanannen da'awar fasaha daga harshen talla.
Kudin da kasafin kuɗi
Kuna iya yin mafi kyawun tambayoyin aiwatarwa kafin kashe kuɗi ko lokaci.
Ƙungiya da aikin aiki
Ƙungiyoyin da ke da fahimtar juna suna yin mafi kyawun samfura, manufofi, da yanke shawara na koyo.
Makomar Wasa Kai Mai Kyau
Wasan kai shine babban ɗan takara don karya bangon bayanai, tunda ta kera nata tsarin karatu maimakon dogaro da ƙarancin lakabin ɗan adam. Yi tsammanin haɓakawa a cikin wuraren da za'a iya tabbatarwa kamar lissafi, lamba, da ka'idar tabbatarwa, inda masu duba atomatik ke tantance yunƙurin samar da kansu. Haɗari sun haɗa da satar lada da rugujewar ƙira daga horo kan kayan aikin roba da yawa, don haka tsarin nan gaba zai iya haɗa wasan kai tare da sigina na ƙasa, masu tabbatarwa, da ra'ayoyin ɗan adam na lokaci-lokaci ko ainihin duniya.
Aiwatar da Gaskiyar Duniya
AlphaGo Zero da AlphaZero sun isa Go, Ches, da Shogi gaba ɗaya ta hanyar wasan kai ba tare da wasannin ɗan adam ba.
SPIN yana haɓaka ƙididdige ma'auni na LLM ta hanyar banbance abubuwan da aka fitar daga amsoshi na ɗan adam.
Samfuran lissafi da ƙididdigewa suna samar da yunƙurin mafita, sannan horarwa akan waɗanda aka tantance ta masu duba atomatik ko gwaje-gwajen rukunin
Wakilan tattaunawa da tattaunawa suna inganta dabarun ta hanyar maimaita wasa biyu na tattaunawar gaba da kansu
Hatsari & Tsare-tsare
Ƙungiyoyi daban-daban na iya amfani da kalmar iri ɗaya daban, don haka ayyana iyaka da wuri.
Alamomi na iya yin kama da ƙarfi yayin da aikin zahirin duniya bai yi daidai ba.
Yin watsi da ingancin bayanai da tsare-tsaren kimantawa galibi yana haifar da sakamako mara ƙarfi.
Taswirar Hanya
Fara da ma'anar harshe a sarari na sakamakon da kuke buƙata.
Zaɓi ma'aunin nasara ɗaya da yanayin gazawa ɗaya kafin gwaji.
Gudun ƙaramin matukin jirgi tare da bayanan wakilci, ba saitin demo da aka goge ba.
Daftarin aiki inda kai-Play Fine-Tuning taimaka da kuma inda mafi sauki hanyoyin ne mafi alhẽri.
Ci gaba da Bincike
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 Self-Play Fine-Tuning 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
Jagora na gaba
Kyakkyawan-Tuning
Tambayoyin da ake yawan yi
What is Self-Play Fine-Tuning?
Daidaita wasa da kai yana inganta samfurin ta hanyar yin gogayya da ko koya daga abubuwan da ya gabata, samar da siginar horo. Yana da mahimmanci saboda yana iya tura aiki fiye da bayanan da ake kulawa ta amfani da ƙaramin ko babu ƙarin lakabin ɗan adam.
Wane sanannen tsarin ne ya isa wasan Go na ɗan adam ta amfani da wasan kai kawai kuma babu bayanan wasan ɗan adam?
AlphaGo Zero ya koyi gabaɗaya daga wasannin da ya buga da kansa, yana farawa daga wasan bazuwar kuma ya zarce nau'ikan da aka horar da su akan wasannin ɗan adam.
A cikin SPIN, menene ke taka rawar 'aboki' wanda dole ne samfurin ya doke?
SPIN tana ɗaukar kyakkyawan daidaitawa azaman wasan wasa kai tsaye inda abubuwan da aka ƙirƙiro da kansu a baya suna aiki azaman abubuwan da samfurin ya koya ya zarce.
Menene babban fa'idar ingantaccen bayanai na wasan-kan-da-kai mai kyau?
Wasan kai yana ƙera siginar horo na kansa, don haka kafaffen saitin kulawa zai iya haifar da ƙarin haɓakawa ba tare da tattara sabbin zanga-zanga ba.
A cikin manufar horon SPIN, menene aka tura samfurin ya yi?
SPIN tana amfani da asara irin ta DPO wanda ke ba da lada mai ban sha'awa na amsoshi na ɗan adam (tabbatacce) daga amsoshi na farko da aka ƙirƙiro (marasa kyau).
Me yasa wahalar yin wasa da kai yana ƙaruwa ta atomatik akan maimaitawa?
Domin kowane ɓacin rai ya fito daga mafi ƙarfi na baya samfurin, ƙirar tana fuskantar ƙalubale mai wuyar ci gaba ba tare da ƙirƙira manhajar karatu ta hannu ba.