Nduzi Asụsụ AI

Usoro nlekọta maka mgbakọ na mwepụ

Nlekọta usoro na-akwụghachi ihe nlereanya maka usoro ọ bụla ziri ezi n'usoro echiche, ọ bụghị naanị azịza ikpeazụ.

2 nkeji na-agụEmelitere ikpeazụ

Nchịkọta

For math, where one wrong move ruins everything, grading the work itself produces far more reliable solvers.

Ime miri emi

Ọtụtụ ụdị ụgwọ ọrụ na-enweta naanị azịza ikpeazụ (nleba anya nsonaazụ). Nke ahụ na-ahapụ ihe nlereanya 'nweta chioma' - iru nọmba ziri ezi site na usoro ndị nwere ntụpọ na-akagbu. Nlekọta usoro na-azụ ụdị ụgwọ ọrụ usoro (PRM) na akara mmadụ ma ọ bụ AI nke na-egosi nzọụkwụ etiti ọ bụla dị ka nke ziri ezi, ezighi ezi, ma ọ bụ na-anọpụ iche. Akwụkwọ 'AIU_PROTECTED_10__'s 2023 'Ka anyị nyochaa nzọụkwụ site na nzọụkwụ' ewepụtara PRM800K, ihe dị ka akara ọkwa ọkwa 800,000 na nsogbu MATH, wee gosi nkwenye na-ahụ maka usoro na-edozi 78% nke ntinye ule na nsonaazụ adịghị ike-naanị ntọala. A na-eji PRM mee ihe n'ịtụle ọtụtụ ngwọta atụpụtara, na-ewere agbụ ahụ nwere akara nrịgo kachasị elu. Ọ na-enyekwa nzaghachi nwere ike ịkọwa: ị nwere ike ịhụ kpọmkwem ebe echiche ahụ gbajiri.

Nghọta nka nka

N'oge ule, ihe nlereanya ahụ na-egosipụta ọtụtụ ngwọta nke onye mmeri; PRM na-enyocha nzọụkwụ ọ bụla yana akara n'ozuzu nke ngwọta bụ ngwaahịa (ma ọ bụ opekempe) nke ihe omume ọ bụla nke ziri ezi. 'Kachasị mma-nke-N' wee họrọ yinye akara kacha elu. N'ihi na a na-ekenye kredit na mpaghara, mgbama ọzụzụ ahụ pere mpe ma dịkwa obere mkpọtụ karịa otu ụgwọ ọrụ ngwụcha nke usoro, nke na-ebelata mbanye ụgwọ ọrụ ebe nzọụkwụ na-ezighi ezi na-enweta azịza ziri ezi.

Mmetụta atụmatụ

Ọsọ na ọnụ ọgụgụ

Usoro ọrụ asụsụ nwere ike ịga ngwa ngwa n'achụghị nkwụsi ike.

Nweta na iru

Ọ na-agbasawanye ohere n'ofe asụsụ na ụdị nzikọrịta ozi.

Mkpebi doro anya

Ndị otu nwere ike itinyekwu oge na ikpe ebe akpaaka na-ejikwa nkwughachi.

Ọdịnihu nke nlekọta usoro maka echiche mgbakọ na mwepụ

Ịde akara akara akwụkwọ ntuziaka dị oke ọnụ, yabụ nyocha na-agbanwe gaa na nlekọta usoro akpaaka - iji Monte Carlo rollouts (Math-Shepherd) tụọ uru nzọụkwụ ọ bụla na-enweghị akara mmadụ, ma ọ bụ inwe ụdị siri ike na-ekpe ndị na-esighị ike. Na-atụ anya PRM ka ha na-emegharị mma-ịmụta nkwalite, ọ bụghị naanị ngbanwe, na ịgbasa karịa mgbakọ na mwepụ n'ime koodu, ihe akaebe sayensị, na nhazi usoro ọtụtụ usoro ebe nhazi ọkwa ọkwa dị mkpa.

Mmejuputa n'ezie n'ụwa

OpenAI's PRM800K dataset: 800K akara ọkwa ọkwa mmadụ ejiri na-azụ ndị nyocha na akara akara MATH

Math-Shepherd: na-akpọ ọkwa ziri ezi na-akpaghị aka site na Monte Carlo rollouts iji zere nkọwa mmadụ dị oke ọnụ.

Ntughari kacha mma nke-N: iwepụta azịza 256 na ịhọrọ nke PRM nwere akara kacha elu na usoro ọ bụla.

Ngwá ọrụ nkuzi na-egosi akara ahịrị ziri ezi na ngwọta nwa akwụkwọ na-arụ ọrụ ebe njehie mbụ pụtara

Ihe ize ndụ & okporo ụzọ nche

Eziokwu ndị e chepụtara echepụta nwere ike jiri nwayọ tinye akụkọ, nkwado nkwado, ma ọ bụ nsonaazụ nyocha.

Mmetụta ngwa ngwa nwere ike ịmepụta nsonaazụ na-ekwekọghị ekwekọ n'ofe arịrịọ ndị yiri ya.

Enwere ike ikpughe data ederede nwere mmetụta ma ọ bụrụ na njikwa ohere adịghị ike.

Map mmejuputa

1

Kọwaa usoro mmepụta, ụda, na ụkpụrụ ịdịmma tupu ibugharị.

2

Weghachite nzaghachi site na isi mmalite ntụkwasị obi mgbe ọ bụla izi ezi dị mkpa.

3

Debe ebe nleba anya mmadụ maka mpụta dị elu.

4

Sochie ụkpụrụ ọdịda ma na-azụghachi mkpali ma ọ bụ usoro ọrụ mgbe niile.

Nọgide na-eme nchọpụta

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 Process Supervision for Math Reasoning quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Malite ajụjụ

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ntuziaka na-esote

Ịtụgharị uche nke Chain-of-Tought

Ajụjụ a na-ajụkarị

What is Process Supervision for Math Reasoning?

Nlekọta usoro na-akwụghachi ihe nlereanya maka usoro ọ bụla ziri ezi n'usoro echiche, ọ bụghị naanị azịza ikpeazụ. Maka mgbakọ na mwepụ, ebe otu mmegharị na-ezighị ezi na-emebi ihe niile, ịdepụta ọrụ n'onwe ya na-arụpụta ihe ngwọta ndị a pụrụ ịdabere na ya karị.

Kedu ihe bụ isi ihe dị iche n'etiti nlekọta usoro na nlekọta nsonaazụ?

Nlekọta usoro na-enye mgbama nkwụghachi ụgwọ n'usoro echiche ọ bụla, ebe nlekọta nsonaazụ na-enyocha azịza ikpeazụ.

Gịnị kpatara na nlekọta naanị nsonaazụ nwere ike bụrụ ihe na-eduhie maka nsogbu mgbakọ na mwepụ?

Na-akwụghachi ụgwọ naanị azịza ikpeazụ na-enweta azịza 'kechioma' ebe usoro etiti etiti na-ezighi ezi na-arụpụta nsonaazụ ziri ezi.

Kedu ihe OpenAI wepụtara n'akụkụ ọrụ 'Ka anyị nyochaa nzọụkwụ site na nzọụkwụ'?

PRM800K nwere ihe dị ka ndetu mmadụ 800,000 na-egosi usoro ntụgharị uche nke onye ọ bụla dị ka nke ziri ezi ma ọ bụ ezighi ezi.

Kedu otu a na-esi ejikarị ụdị nkwụghachi ụgwọ usoro n'oge ntinye aka?

PRM na-enyocha usoro nke ọ bụla nke ọtụtụ ngwọta ndị ndoro-ndoro anya, na-eme ka nhọrọ kacha mma nke-ntụgharị uche dị elu.

Kedu ụzọ na-ebelata mkpa maka akara nzọụkwụ mmadụ dị oke ọnụ?

Math-Shepherd na-atụle izi ezi nzọụkwụ site n'iwepụta mmecha yana tụọ ugboro ole ha ruru azịza ziri ezi, na-ezere ịde aha akwụkwọ ntuziaka.