Online platforms usually suffer from user experience degradation due to near-duplicate items with similar visuals and text. While Multimodal Large Language Models (MLLMs) enable multimodal embedding, existing methods treat representations as black boxes, ignoring structural relationships (e.g., primary vs. auxiliary elements), leading to local structural collapse problem. To address this, inspired by Feature Integration Theory (FIT), we propose FITRep, the first attention-guided, white-box item representation framework for fine-grained item deduplication. FITRep consists of: (1) Concept Hierarchical Information Extraction (CHIE), using MLLMs to extract hierarchical semantic concepts; (2) Structure-Preserving Dimensionality Reduction (SPDR), an adaptive UMAP-based method for efficient information compression; and (3) FAISS-Based Clustering (FBC), a FAISS-based clustering that assigns each item a unique cluster id using FAISS. Deployed on Meituan's advertising system, FITRep achieves +3.60% CTR and +4.25% CPM gains in online A/B tests, demonstrating both effectiveness and real-world impact.

I found that the reality of drone delivery is still far from ideal, and people may be turned away by the steep learning curve. But at the same time, it was an exciting experience--the prospect of routine drone delivery feels more realistic than it's ever been. Meituan currently operates more than a hundred drones from five delivery hubs (or launchpads) in the city. Together, they completed over 100,000 orders in 2022. While the platform itself can deliver basically anything, from dinner to medicine to fresh flowers to electronic devices, the drones are mostly used for food and drinks.

The drone delivery service I was trying out is operated by Meituan, China's most popular food delivery platform. In 2022, the company engaged some 6 million gig delivery workers to deliver billions of orders. But the company has also been developing drone delivery since 2017. And in Shenzhen, a southern city that's home to a mature drone supply chain, Meituan has been regularly operating such delivery routes for the last year and a half. Many big corporations have had their eyes on drone delivery: Amazon first proposed doing it in 2013, but its progress has been limited by regulations and a lack of demand.

Nowadays, the mainstream approach in position allocation system is to utilize a reinforcement learning model to allocate appropriate locations for items in various channels and then mix them into the feed. There are two types of data employed to train reinforcement learning (RL) model for position allocation, named strategy data and random data. Strategy data is collected from the current online model, it suffers from an imbalanced distribution of state-action pairs, resulting in severe overestimation problems during training. On the other hand, random data offers a more uniform distribution of state-action pairs, but is challenging to obtain in industrial scenarios as it could negatively impact platform revenue and user experience due to random exploration. As the two types of data have different distributions, designing an effective strategy to leverage both types of data to enhance the efficacy of the RL model training has become a highly challenging problem. In this study, we propose a framework named Multi-Distribution Data Learning (MDDL) to address the challenge of effectively utilizing both strategy and random data for training RL models on mixed multi-distribution data. Specifically, MDDL incorporates a novel imitation learning signal to mitigate overestimation problems in strategy data and maximizes the RL signal for random data to facilitate effective learning. In our experiments, we evaluated the proposed MDDL framework in a real-world position allocation system and demonstrated its superior performance compared to the previous baseline. MDDL has been fully deployed on the Meituan food delivery platform and currently serves over 300 million users.

Chinese artificial intelligence chipmaker Axera Technology has raised CNY800 million (USD126 million) from renowned investors such as food delivery giant Meituan in its second round of financing within the past six months. Several investment institutions, including Qiming Venture Partners and GGV Capital, as well as Beijing-based Meituan joined Axera's Series A round of funding, its fourth fundraiser, the Beijing-headquartered startup said in a statement on its WeChat account today. The firm intends to use the proceeds to attract top industry talent and to expand its business, it added. The global demand for AI chips is growing explosively, said Axera. The market scale is expected to grow to USD72.6 billion by 2025 after reaching USD17.5 billion last year, according to data from AI research institute Tractica.

While the past 60 years have defined the field of industrial robots and empowered hard-bodied machines to execute complex assembly tasks in constrained industrial settings, the next sixty years will usher in robots in human-centric environments to assist humans with physical tasks. While the industrial robots of the past 60 years have mostly been inspired by the human form, the next stage will be soft robots inspired by the animal kingdom: form and diversity modeled by our own built environment, with broad potential to mimic our natural state.

On a congested sidewalk next to a busy mall in Shenzhen, a 20-something woman uses a smartphone app to order a milk tea on Meituan, a major food delivery company. In under ten minutes, the pearl-white drink arrives, not on the back of one of the city's ubiquitous delivery bikes, but descending from the cloudy heavens, in a cardboard box on the back of a drone, into a small roadside kiosk. The only thing the scene is missing is a choir of angels. Over the past two years, Meituan, one of China's largest internet companies, has flown 19,000 meals to 8,000 customers across Shenzhen, a city with close to 20 million people. The pilot program is available to just seven neighborhoods, each with a three-kilometer stretch, and only from a select number of merchants.

More than a thousand robots are set to join the delivery personnel ranks of Chinese behemoths Alibaba, Meituan and JD.com over the next year as the pandemic fuels demand for contactless services. The firms expect to operate over 2,000 robots between them by 2022, up about fourfold from now, their executives said, encouraged also by falling costs of making robots. Millions of couriers still deliver packages for as less as 3 yuan ($0.47) in China, but companies have been exploring the use of drones or box-like robots on wheels from as early as 2013 amid a labor crunch that has worsened due to the pandemic. Beijing has also ordered firms to ensure rest periods for couriers as they scramble to meet rising demand and deadlines. "The COVID-19 pandemic has been a big boost" for robot rollout plans, said Xia Huaxia, chief scientist at Meituan.

Beijing's crackdown on its tech giants is fueling a noticeable phenomenon: it's opened the spending floodgates. China's largest internet corporations are digging deep into their pockets to open up new avenues of growth as Beijing curtails their most lucrative businesses from fintech to e-commerce. Tencent Holdings Ltd., Alibaba Group Holding Ltd. and Meituan have all warned investors in recent weeks they're prepared to open their coffers to expand in areas such as cloud computing, autonomous driving and artificial intelligence. The coming deluge promises to transform the internet landscape by funneling capital into fundamental technology and infrastructure -- not coincidentally priority areas for the Communist Party. Of the three, Beijing-based Meituan was the earliest to throw caution to the wind and also the one most accustomed to sacrificing profits.

The Hong Kong-listed company is selling as much as $7 billion of stock and $3 billion of convertible bonds, according to a term sheet Monday that was seen by The Wall Street Journal. The huge capital-raising is considerably bigger than the $4.2 billion Meituan raised in its 2018 initial public offering and suggests there is still a healthy appetite among investors for stock in Chinese technology companies, even though shares in Meituan and many peers have pulled back recently. Earlier this month, Prosus NV raised $14.6 billion by selling down a small part of its stake in Tencent Holdings Ltd., the internet and videogaming giant. Meituan plans to use some of the net proceeds on projects such as researching and developing autonomous delivery vehicles, drone deliveries and other cutting-edge technology, the term sheet said. In a separate statement Monday, the company unveiled a new generation of self-driving delivery cars that it said were smarter and safer than previous versions, with a longer battery life and capable of carrying heavier loads.