Vision-Language Pre-training (VLP) models have shown remarkable performance on various downstream tasks. Their success heavily relies on the scale of pre-trained cross-modal datasets. However, the lack of large-scale datasets and benchmarks in Chinese hinders the development of Chinese VLP models and broader multilingual applications. In this work, we release a large-scale Chinese cross-modal dataset named Wukong, which contains 100 million Chinese image-text pairs collected from the web. Wukong aims to benchmark different multi-modal pre-training methods to facilitate the VLP research and community development.

The latest addition to China's Tiangong space station is an AI chatbot with expertise in navigation and tactical planning. Named Wukong AI--after the protagonist of the "Monkey King" legend in Chinese mythology, Sun Wukong--the chatbot was introduced on the space station in mid-July, and has already completed its first mission: supporting three taikonauts during a spacewalk. Information about Wukong AI remains limited. Chinese authorities have said that they developed it from a domestic open-source AI model; according to Xinhua, China's state-run news agency, engineers designed it to meet the requirements of manned space missions, and focused its knowledge-base on aerospace flight data. "This system can provide rapid and effective information support for complex operations and fault handling by crew members, improving work efficiency, in-orbit psychological support, and coordination between space and ground teams," Zou Pengfei of the taikonaut training center, told Xinhua.

It's easy to feel a bit beset by doom these days. The other week, I watched the heinous AI-generated "Trump Gaza" video and was so appalled that I impulse-bought a kayaking guide book. It felt like the only sane response was to take to the water and paddle away. Video games are a reliable antidote to existential doom, but layoffs, corporate homogenisation and AI slop are all encroaching on my safe haven, making it more difficult to get a brief reprieve from what's happening in the outside world. Thank God, then, for the Bafta games awards nominations, which reliably remind me that video games are pretty great, actually.

Vision-Language Pre-training (VLP) models have shown remarkable performance on various downstream tasks. Their success heavily relies on the scale of pre-trained cross-modal datasets. However, the lack of large-scale datasets and benchmarks in Chinese hinders the development of Chinese VLP models and broader multilingual applications. In this work, we release a large-scale Chinese cross-modal dataset named Wukong, which contains 100 million Chinese image-text pairs collected from the web. Wukong aims to benchmark different multi-modal pre-training methods to facilitate the VLP research and community development.

Much like Christmas is a lot less enjoyable for the person who has to organise all the presents and cook the dinner, game-of-the-year season is rather intimidating for the people who have to put together the shortlists. Every November, I tot up all of the year's acclaimed games I've yet to play, the underground recommendations I've yet to follow up on and the games I loved back in February but forgot about. I feel a mounting panic. And when all of the year-end lists come out, I inevitably find I've missed something anyway. The Game Awards have just announced the nominations for this year's ceremony, taking place on 12 December in Los Angeles.

A new Chinese video game has created a buzz worldwide after it sold more than 10 million copies within three days, becoming the most successful game of all time to emerge from the country. According to 2023 estimates, China's gaming industry is roughly worth 40bn. Black Myth: Wukong, produced by developer Game Science (GS), has already generated an estimated 800-900m in revenue to date and will help project Chinese culture to a global audience. The game, believed to be China's first AAA video game, was developed at a reported cost of about 70m over six years. AAA is a classification used to denote a high-budget or high-profile game from a large video game developer.

A Chinese game called Black Myth: Wukong has been the biggest hit of the summer, selling 10m copies in just three days, according to its developer Game Science, with over 1 million people playing it every day on games marketplace Steam. China's homegrown games industry is absolutely massive, but concentrated almost entirely on mobile phones: this is the country's first successful blockbuster console and PC game, which makes it very interesting in itself. It's also a massively successful single-player game arriving on the back of a few high-profile multiplayer flops, which suggests there is still more of a market for this kind of adventure than video game execs like to believe. But Wukong has been grabbing headlines for other reasons, too. Back in November, IGN put together a report compiling crude, vulgar public comments from a number of Game Science staff, some of whom are very well-known in China's games industry.

When Chinese developer Game Science revealed its debut console game Black Myth: Wukong last year, it immediately caused a stir. Inspired by the great 16th-century Chinese novel, Journey to the West, the action-packed footage featured the titular mythological monkey Sun Wukong battling Buddhist-folklore demons and sword-wielding anthropomorphic foxes in lusciously rendered forests. Smartphone games are inordinately popular in China, but console game developers are still few and far between, and the excitement for Wukong in Game Science's homeland reached fever pitch. Within 24 hours, the trailer racked up 2m views on YouTube and more than 10m on Chinese video sharing site Bilibili, much to its creators' shock and delight. One excited fan even broke into the developer's office, desperate for more info on the game.

Scaling laws play an instrumental role in the sustainable improvement in model quality. Unfortunately, recommendation models to date do not exhibit such laws similar to those observed in the domain of large language models, due to the inefficiencies of their upscaling mechanisms. This limitation poses significant challenges in adapting these models to increasingly more complex real-world datasets. In this paper, we propose an effective network architecture based purely on stacked factorization machines, and a synergistic upscaling strategy, collectively dubbed Wukong, to establish a scaling law in the domain of recommendation. Wukong's unique design makes it possible to capture diverse, any-order of interactions simply through taller and wider layers. We conducted extensive evaluations on six public datasets, and our results demonstrate that Wukong consistently outperforms state-of-the-art models quality-wise. Further, we assessed Wukong's scalability on an internal, large-scale dataset. The results show that Wukong retains its superiority in quality over state-of-the-art models, while holding the scaling law across two orders of magnitude in model complexity, extending beyond 100 GFLOP/example, where prior arts fall short.

Every August for the past two years, Chinese developer Game Science Studio has released new gameplay footage from its upcoming action RPG Black Myth: Wukong. Not one to miss a beat, it has done the same this year. On Friday, the studio shared a new eight-minute gameplay trailer and six-minute in-game cutscene. Much like last year's Unreal Engine 5 reveal, the former is partly a showcase for NVIDIA's DLSS AI-powered upscaling tech, and you can see what a difference it – and a year of additional work – has meant for the game's framerate. Compared to last year's trailer, the action is smoother and there are fewer framerate drops. We also see Game Science Studio iterate on From Software's Souls formula in a few interesting ways.