NVIDIA, Bulgaria’s INSAIT develop platform to evaluate AI Models in robotics
Baku, October 26, AZERTAC
Tech giant NVIDIA and the Institute for Computer Science, Artificial Intelligence and Technology (INSAIT) at Sofia University have developed a platform called LADDER that allows researchers to assess and compare the capabilities of leading artificial intelligence (AI) models in controlling robots as they perform sequences of actions, the Bulgarian News Agency (BTA) reported citing the Ministry of Education and Science's press centre.
Although AI can now solve complex mathematical problems and humanoid robots can perform impressive movements, controlling robots as they carry out everyday tasks remains a major scientific challenge. Moving a single object is a relatively easy task, but when a robot has to move several objects in sequence, the likelihood of errors increases significantly.
The project focuses on one of the most promising areas of modern technology: Physical AI, in which AI systems must not only process information but also interact with the real world.
NVIDIA is one of the world's leading tech giants, with its developments underpinning the AI revolution. The company is a major player in the development of the most powerful AI models, while robotics and Physical AI are among its strategic priorities. It was in this field that research by Bulgaria's INSAIT attracted the attention of NVIDIA, which recognized its potential and joined the effort to develop it further. The result of this collaboration is the LADDER platform, which has the potential to become an important benchmark for assessing progress in robotics and the development of the next generation of intelligent robots. The partnership is significant recognition of INSAIT's research potential and its standing among international research centres, the press release noted.
The initial results show that even the most powerful systems are still far from being able to control robots reliably. The leading models in the rankings include GPT-6 Astra, Claude Opus 5.5 and Gemini 4 Argon. Although they outperform other models, they fail to complete a substantial proportion of the tasks. In the most challenging tests, which involve eight to ten consecutive actions, even the best-performing models succeed in only about one-third of attempts. Models specifically designed for robot control face even greater difficulties. One of the leading models in this category can move a single object but almost never manages to complete five or more consecutive actions.
The study also identifies another significant challenge: the high cost of using the most powerful AI models. For some models, the average cost of successfully completing a single task exceeds USD 6, while a full evaluation involving nearly 2,000 attempts would cost around USD 8,000.
The platform is publicly accessible and will be updated with the results of new AI models. The rankings, videos showing the tasks being performed and the full research paper are available at https://ladderrobotics.ai.
On October 23, Google made a new USD 1 million donation to INSAIT. It marks another step in the strategic partnership between Google and the Bulgarian research institute. Google's total support for the Bulgarian institute has now exceeded USD 7 million, an unprecedented level of backing for Bulgaria and the region, the university noted.
INSAIT is the first institution in Eastern Europe to offer a world-class research environment on a par with those at leading Western universities and research institutes. It is established as a dedicated unit within Sofia University and was founded in partnership with Switzerland's ETH Zurich and EPFL, two of the world's leading technical universities. Prominent scientists from top universities and research laboratories in the United States, Europe and Israel lead and advise the institute. INSAIT's main focus is on scientific excellence: conducting world-class research, attracting internationally renowned scientists with outstanding achievements, and educating the next generation of students and technology leaders.
Other news in this section
