HAGAHA Farsamada

Tusaalooyinka Cadawga iyo Adkaynta

Tusaalooyinka iska soo horjeeda waa wax-soo-saarka ay xumeeyaan isbeddellada yaryar, inta badan aan la fahmi karin kuwaas oo keena moodal si kalsooni leh, saadaalin khaldan.

2 daqiiqo akhriMarkii u dambaysay ee la cusbooneysiiyay

Dulmar

Adkeygu waa goobta u heellan in laga difaaco iyaga, waxayna muujineysaa farqiga qoto dheer ee u dhexeeya mashiinka iyo aragtida aadanaha.

quusid qoto dheer

2013-2014, cilmi-baarayaashu waxay muujiyeen in ku darista si taxadar leh loo sameeyay, qaab dhawaqa aan la arki karin ee sawirka ay ka rogi karto kala-soocida 'panda' ilaa 'gibbon' kalsooni sare leh. Tusaalooyinkan iska soo horjeeda waxay ka faa'iidaystaan ​​xaqiiqda ah in shabakadaha neural ay bartaan xuduudaha go'aannada ee jajaban ee meel sare. Weeraradu sida caadiga ah waa sanduuqa cad (weerarahu wuu yaqaan qaabka oo wuxuu isticmaalaa gradients, sida FGSM iyo PGD) ama sanduuqa madow (kaliya wax soo saarka ayaa muuqda). Waxa cajiib ah, tusaalooyinka iska soo horjeeda inta badan waxay u kala wareejiyaan noocyo kala duwan, taas oo awood u siinaya weerarrada iyada oo aan gudaha la gelin. Khatartu waa mid wax ku ool ah: dhejisyada muuqa-adduunku waxay khiyaami karaan baarayaasha joogsiga joogsiga, iyo duritaanka degdega ah ee 'jailbreaks' waa analoogga qaabka luqadda. Cilmi-baarista adkeysigu waxay raadisaa moodallo si sax ah u dhaqma xitaa marka ay xaaladdu ugu xun tahay, dhibka iska soo horjeeda.

Aragtida Farsamada

Weeraro badan ayaa ku salaysan gradient: FGSM waxay qaadaysaa hal tallaabo jihada calaamadda jaan-goynta luminta iyadoo la eegayo gelinta, halka PGD ay tan ku cabbirayso kubad xaddidan oo yar (tusaale, L-infinity) oo ku wareegsan gelinta asalka ah. Difaaca ugu xooggan ee la yaqaan waa tababbarka iska hor imaadka, dib u tababbarka tusaalooyinka iska soo horjeeda, oo loo qaabeeyey dhibka ugu yar: yaree khasaaraha ka dhanka ah dhibka ugu xun. Waxay hagaajinaysaa adkaanta laakiin caadi ahaan waxay ku kacaysaa saxnaanta nadiifka ah iyo xisaabinta.

Saamaynta Istiraatijiyadeed

Qiimaha iyo miisaaniyada

Go'aamada qaab-dhismeedku waxay horseedaan waxqabadka iyo kharashka hawlgalka sannadaha.

Go'aamo cad

Waxbarashada farsamada waxay ka caawisaa kooxaha inay doortaan xidhmo sax ah, ma aha oo kaliya kan ugu cusub.

Xakamaynta tayada

Doorashooyinka injineernimada ee wanaagsan waxay yareeyaan shilalka la isku halleyn karo ee wax soo saarka.

Mustaqbalka Tusaalooyinka Cadawga ah iyo Adkaynta

Marka ay AI soo gasho nidaamyada badbaadada-muhiimka ah, adkaanta ayaa ka guuraysa xiisaha tacliinta una guuraysa shuruudaha injineernimada. Shaqadu waxay ku sii socotaa difaacyo shahaadaysan oo xisaab ahaan dammaanad qaadaya in aanay jirin wax qas ah oo ka jira xuduudda oo beddeli karta wax soo saarka, iyo adkaysiga ka dhanka ah ballaadhan, weerarrada adag ee xidhxidhan ee wajahaya moodooyinka luqadaha waaweyn, sida jeelasha iyo duritaanka degdega ah. Filo halbeegyada iska soo horjeeda, dhuumaha kooxda cas, iyo cadaadiska nidaaminta moodooyinka lagu diro wadista iskeed u madax bannaan, amniga, iyo daryeelka caafimaadka si loo muujiyo isku halaynta xaaladda ugu xun.

Dhaqangelinta Adduunka-dhabta ah

Cilmi-baadhayaashu waxay dhejiyeen dhejisyo yaryar oo jireed calaamad joogsi ah taas oo keentay qaabka aragga inuu si khaldan u akhriyo calaamadda xawaaraha-xawaaraha, taas oo muujinaysa khatarta dhabta ah ee aduunka ee baabuurta iswada.

Kooxaha nabad-gelyada ayaa aqoonsi weji koox-cas ah leh oo leh balastarro iska soo horjeeda oo ku daabacan muraayadaha ama dharka ka baxsanaya ama nacasnimada isbar-barashada aqoonsiga.

Shaandhooyinka spamka iyo malware-ka waxa lagu baadhayaa agab si xun u qalloocan kuwaas oo ilaalinaya culaysyada mushaharka xun marka ay siibtaan kalasaarayaal hore.

Horumarinta LLM waxay ka difaacaan cirbadeynta degdega ah ee 'jailbreaks', luqadda analooga ah ee tusaalooyinka iska soo horjeeda, kuwaas oo moodooyinka khiyaanada ah si ay u indho-tiraan tilmaamaha badbaadada.

Khatarta & Dariiqyada Ilaalada

Hagaajinta hal bartilmaameed waxay qarin kartaa daciifnimada nidaamka ballaaran.

Kaabayaasha dhaqaalaha iyo dayactirka inta badan waa la dhayalsadaa.

Nabadgelyada iyo daldaloolada u fiirsashada ayaa kori kara marka nidaamyadu noqdaan kuwo aad u adag.

Qorshe Hawleedka Dhaqangelinta

1

Qeex daahida, tayada, iyo bartilmaameedyada qiimaha ka hor inta aan la hirgelin.

2

Benchmark marka la eego culeyska dhabta ah iyo xaaladaha xogta.

3

La socodka qalabka khaladaadka, leexashada, iyo saamaynta isticmaalaha.

4

U diyaari dib-u-noqoshada iyo dariiqyada jawaab-celinta dhacdada ka hor inta aanad miisaan.

Sii wad Sahaminta

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 Adversarial Examples and Robustness quiz

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

Bilow kedis

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

Hagaha xiga

Shabakadaha Adversarial Generative

Su'aalaha soo noqnoqda

Waa maxay Tusaalooyinka Cadawga iyo Adkaynta?

Tusaalooyinka iska soo horjeeda waa wax-soo-saarka ay xumeeyaan isbeddellada yaryar, inta badan aan la fahmi karin kuwaas oo keena moodal si kalsooni leh, saadaalin khaldan. Xoognimadu waa goobta loogu talagalay in laga difaaco iyaga, waxayna muujineysaa farqiga qoto dheer ee u dhexeeya mashiinka iyo aragtida aadanaha.

Waa maxay tusaale iska soo horjeeda?

Tusaalooyinka iska soo horjeeda waxay ku daraan qalqal yar yar, si taxadar leh loo farsameeyey oo aadanuhu si dhib yar u ogaan laakiin ku nacaseeya qaabka khaladaadka kalsoonida sare leh.

Maxaa ka sooca weerarka 'sanduuqa cad' iyo weerarka 'sanduuqa-madow'?

Weeraryahannada sanduuqa cad waxay yaqaaniin qaabka waxayna isticmaali karaan jaangooyooyinkooda; Weeraryahannada sanduuqa madow waxay arkaan kaliya wax-soo-saarka iyo soo-saarka.

Waa maxay difaaca ugu badan ee loo isticmaalo wanaajinta adkeysiga iska soo horjeeda?

Tababarka iska horimaadku wuxuu ku kordhiyaa barashada xaalad-xumada ugu xun, oo loo qaabeeyey sida ugu wanagsan ee ugu yar, waana difaaca ugu adag ee la isku halayn karo, inkasta oo ay yarayn karto saxnaanta nadiifka ah.

Waa maxay sababta 'wareejinta' tusaalooyinka iska soo horjeeda ay khuseeyaan?

Sababtoo ah tusaalooyinka iska soo horjeeda waxay u kala wareejiyaan moodooyinka, weeraryahanku wuxuu ku farsamayn karaa qaab beddelka wuxuuna si guul leh u weerari karaa bartilmaameed aysan baari karin.

Waa maxay tusaalaha-qaab-afka-weyn ee tusaalooyinka iska soo horjeeda?

Jeexitaanka iyo duritaanka degdega ah ayaa ah qalab la farsameeyey oo u maamula LLM-ka dhaqan aan badbaado lahayn ama aan la qorshayn, oo barbar socda weerarro iska soo horjeeda oo aragga ah.