Evaluating AI Agent Skill Performance with NVIDIA SkillEvaluator
NVIDIA SkillEvaluator measures the impact of verified skills on AI agent performance through a three-tier evaluation process.Soma inkuru
Inkomoko yemejwe: developer.nvidia.comKuvugururwa buri munsi2273 inkuru zagenzuwe
Amakuru ya AI. Nta rusaku.
Inkomoko yagenzuwe na AI yerekana ibicuruzwa, guhindura politiki, ubushakashatsi ku mutekano, hamwe n’inganda zigenda, byasobanuwe mu Cyongereza cyoroshye nitsinda ry’uburezi ridaharanira inyungu.
Amasoko yagenzuwe
Buri nkuru ihuza ibimenyetso bikomeye biboneka: inkomoko yumwimerere iyo iboneka, naho ubundi raporo yatanzwe neza.
Icyongereza
Ibyabaye, impamvu bifite akamaro, nicyo kureba - nta jargon.
Nta wuzuza
Iyo ikimenyetso ari gito, ntacyo dusohora aho gukanda ibiryo.
Izindi nkuru
9 inkuruGuhanga udushya
Benchmark Says Top Multimodal Models Score Under 10% at Reading Words From Pen Sounds and Hand Motion
A new arXiv paper introduces a test in which models must infer a written word from pen-scratch audio and hand-movement video, with no ink visible. The authors report humans above 80% ordered letter accuracy and leading models below 10% — and that giving models both modalities often made results worse.arxiv.orgGuhanga udushya
Impapuro zivuga ko Abakozi-Bazi Cache Ubuyobozi Bugabanya Gutinda-Bwa mbere Gutinda Kugera kuri 45% muri Serivisi nyinshi
Icapiro rishya rya arXiv risobanura CacheScout, urwego rwubatswe ku isoko ifunguye-vLLM ya seriveri ihitamo icyo ugomba kubika muri cache yingenzi-agaciro kashe ukurikije umukozi ushobora kuzakurikiraho. Abanditsi batangaza imibare ibiri yubukererwe ninyungu zinjira; imizigo y'akazi, icyitegererezo, hamwe nibikoresho ntibivugwa muri abstract.arxiv.orgGuhanga udushya
Audit of an OpenAI AI-Generated Proof Finds a Reversed Condition, and Publishes a Repair
Two researchers say a lemma proof in Chapter 6 of OpenAI's mathematics document has a polarity error: a test in terms of average success where the next step needs a large conditional failure. They give a counterexample and a corrected proof, and caution that this is not verification of the chapter's main theorem.arxiv.orgGuhanga udushya
Replication Study Says FLOPs Still Mispredict AI Runtime, and the Proposed Fix Fails on Newer Hardware
A preprint by two researchers reproduces an earlier study on why equal FLOP counts do not mean equal execution time. It confirms the underlying claim but reports that the α-FLOPs correction formula generally underestimates runtime on newer hardware, which shows jumps and oscillations the formula does not capture.arxiv.orgUruganda
Benchmark Paper Finds Four Ways to Query Enterprise Data With LLMs All Score Under 26%
A new arXiv preprint pits four architectures for natural-language querying of enterprise databases against each other on a synthetic bilingual benchmark. None answered more than about a quarter of cases correctly, and the design that scored highest was not the safest or the cheapest.arxiv.orgGuhanga udushya
Impapuro zisaba gutanga amanota yumutekano wa AI udafite ibirango mugupima guhuza icyitegererezo gikomeye
Icapiro rishya rya arXiv rivuga ko itsinda ry’umutekano rishobora gusuzuma niba umukozi wa AI ufite ibikoresho byo kwibuka cyangwa kugarura ibikoresho yiga mu gupima intera ikuraho icyuho cy’icyitegererezo "umwarimu" ukomeye, aho gushingira ku bipimo byanditseho bikunze kuba bike cyangwa bishaje. Urebye icyitegererezo gisa nacyo nta kimenyetso cyakoreshwa.arxiv.orgGuhanga udushya
New Benchmark Tests Whether AI Assistants Can Remember a Year of Phone Use
A 17-author technical report posted to arXiv introduces MobileMem, a benchmark and framework for on-device long-term memory built from a year-scale collection of mobile experiences. The abstract describes the design but reports no scores, and key details about the underlying data remain undisclosed.arxiv.orgGuhanga udushya
Paper Reports Brain-Like Modular Organization Emerging Inside Large Language Models
A new arXiv preprint says large language models develop functionally specialized internal structure that lines up with distinct human brain networks, based on circuit analyses across 46 tasks in four cognitive domains. The abstract page leaves key methodological details unstated.arxiv.orgGuhanga udushya
Impapuro Zisanga Ibice Byatinze Byivanze-by-Impuguke Model Yihanganira Masking Impuguke
Icapiro rivuga ko kwanga impuguke zifite uburemere buke mu byiciro bitanu byanyuma bya miliyari 35-ibipimo byerekana imvange-y’impuguke yazigamye ibisubizo byifashishwa mu gusobanura-ibisobanuro kuruta gukwirakwiza ibice bimwe mu bice byose. Ikubiyemo icyitegererezo kimwe nigipimo kimwe, na raporo zidafatika nta shingiro zifatika.arxiv.org
Imyitozo imwe y'ingirakamaro buri cyumweru
Komeza hamwe na AI utabayeho mubiryo.
Shakisha icyumweru cyagenzuwe amakuru ya AI, amakuru yumwimerere, ibikoresho byingirakamaro, gutoranya imyigire, nakazi gashya ka AI.
Shikira abantu biga AI
Guha akazi umwuga wa AI cyangwa gutangiza ibicuruzwa bya AI bifite akamaro? Shyira imbere yabantu baje hano kwiga no gukora.
Kohereza akazi ka AITanga igikoresho cya AI