Author

Meituan Technology Team

Over 10,000 engineers powering China’s leading lifestyle services e‑commerce platform. Supporting hundreds of millions of consumers, millions of merchants across 2,000+ industries. This is the public channel for the tech teams behind Meituan, Dianping, Meituan Waimai, Meituan Select, and related services.

646

Articles

Likes

Views

Comments

Latest from Meituan Technology Team

100 recent articles max

Meituan Technology Team

Mar 6, 2025 · Artificial Intelligence

INT8 Quantization and Inference Optimization of DeepSeek R1 Model

Meituan’s search and recommendation team converted the FP8‑only DeepSeek‑R1 model to INT8 by first casting weights to BF16 and then applying block‑wise or channel‑wise quantization, which preserves GSM8K and MMLU accuracy while delivering 33% to 50% higher throughput on A100‑80G GPUs, and they released the SGLang‑based inference scripts and quantized weights publicly, enabling deployment on older NVIDIA hardware without accuracy loss.

DeepSeek-R1GPU deploymentINT8 quantization

0 likes · 11 min read

INT8 Quantization and Inference Optimization of DeepSeek R1 Model

Meituan Technology Team

Feb 27, 2025 · Artificial Intelligence

ECUP and NLGR: Context-Aware Uplift Modeling and Reranking for Meituan Aggregation Page Ads

The paper introduces ECUP, a context‑enhanced uplift‑modeling framework that mitigates chain bias and treatment mismatch through a full‑chain enhancement network, task‑enhanced priors, and bit‑level treatment adaptation, achieving superior AUUC and QINI scores and online A/B gains for Meituan’s coupon issuance, and NLGR, a neighbor‑list generative reranking system that leverages non‑autoregressive sampling and reward‑based training to boost hit‑ratio performance on public and internal datasets, demonstrating the effectiveness of context‑aware uplift modeling and neighbor‑list reranking for aggregation‑page advertising.

Context-Aware LearningMeituanNeighbor Lists

0 likes · 14 min read

ECUP and NLGR: Context-Aware Uplift Modeling and Reranking for Meituan Aggregation Page Ads

Meituan Technology Team

Feb 20, 2025 · Artificial Intelligence

Offline Multi-Agent Reinforcement Learning via In‑Sample Sequential Policy Optimization (InSPO)

Offline multi‑agent reinforcement learning (MARL) faces challenges such as out‑of‑distribution joint actions and local optima, and this article introduces the In‑Sample Sequential Policy Optimization (InSPO) algorithm—leveraging inverse KL regularization, maximum‑entropy, and cooperative Markov games—to achieve monotonic policy improvement and superior performance across benchmark tasks.

InSPOMaximum Entropycooperative Markov game

0 likes · 18 min read

Offline Multi-Agent Reinforcement Learning via In‑Sample Sequential Policy Optimization (InSPO)

Meituan Technology Team

Feb 13, 2025 · Cloud Computing

Predictive Auto-Scaling System (PASS) for Large-Scale Enterprise Web Applications

The Predictive Auto‑Scaling System (PASS) jointly developed by Meituan and Prof. Chai Yunpeng’s team uses an ensemble learning‑based prediction algorithm that dynamically selects and calibrates online and offline models, combined with a log‑derived performance model and a hybrid scaling strategy, to accurately forecast QPS bursts, proactively adjust resources, and achieve up to 40 % cost savings while maintaining strict QoS guarantees for large‑scale enterprise web services.

Auto ScalingPredictive Modelingcloud computing

0 likes · 23 min read

Predictive Auto-Scaling System (PASS) for Large-Scale Enterprise Web Applications

Meituan Technology Team

Feb 9, 2025 · Artificial Intelligence

NTIRE 2025 XGC AI-Generated Video Quality Assessment Challenge

The NTIRE 2025 XGC AI‑Generated Video Quality Assessment Challenge, hosted at the CVPR workshop, invites participants to build VQA models that predict mean opinion scores for 34,029 AI‑generated videos created from 4,689 prompts using 14 generation models, with training, validation, and test splits provided as JSON, and submissions evaluated by the average of PLCC and SROCC, while key dates run from February 5 to June 15 2025 and prize money up to $1,200 is offered.

AI videoCVPRChallenge

0 likes · 6 min read

NTIRE 2025 XGC AI-Generated Video Quality Assessment Challenge

Meituan Technology Team

Jan 9, 2025 · Artificial Intelligence

Roundtable Discussion on Embodied Intelligence at Meituan Robot Research Institute 2024 Academic Annual Meeting

At Meituan Robot Research Institute’s 2024 academic meeting, a diverse panel of scholars and entrepreneurs debated the relative importance of hardware and algorithms for embodied intelligence, identified near‑term market niches such as hazardous‑environment and household assistance, projected rapid scaling to thousands of autonomous humanoids, and highlighted safety, mass‑market adoption, and ethical considerations as key challenges.

Artificial IntelligenceIndustry ApplicationsRobotics

0 likes · 27 min read

Roundtable Discussion on Embodied Intelligence at Meituan Robot Research Institute 2024 Academic Annual Meeting

Meituan Technology Team

Jan 2, 2025 · Artificial Intelligence

Meituan Robot Research Institute 2024 Academic Annual Meeting and Low‑Altitude Flight Challenge

At its December 27, 2024 academic annual meeting in Shenzhen, Meituan Robot Research Institute showcased a year of progress in embodied intelligence and large‑model robotics, honored 109 university teams from 74 schools in the 2nd Low‑Altitude Economy Intelligent Flight Management Challenge, highlighted collaborations with more than 20 top universities that yielded over 40 joint projects and papers at premier conferences, announced a city‑wide digital‑twin low‑altitude simulation using real drone data, and outlined future plans for expanding the challenge platform, offering internships, creating innovation scholarships, and inviting external research proposals, while a round‑table of leading experts discussed commercialization prospects.

Artificial IntelligenceLow‑Altitude FlightRobotics

0 likes · 6 min read

Meituan Robot Research Institute 2024 Academic Annual Meeting and Low‑Altitude Flight Challenge

Meituan Technology Team

Jan 2, 2025 · Information Security

Analysis of HarmonyOS NEXT Application Signature and Verification Mechanism

The article thoroughly examines HarmonyOS NEXT’s application signing and verification process, detailing how developers generate PKCS#12 certificates, create CSRs and leaf certificates, embed a signed Profile in a custom HapSigningBlock within the .app/.hap ZIP, and how the OS at runtime validates the certificate chain, profile authenticity, package integrity, and distribution constraints, thereby combining Android‑style CMS signatures with iOS‑like provisioning to ensure robust app security.

App SigningHarmonyOSOpenHarmony

0 likes · 33 min read

Analysis of HarmonyOS NEXT Application Signature and Verification Mechanism

Meituan Technology Team

Nov 21, 2024 · Mobile Development

AutoConsis: Automated UI Consistency Detection for Mobile Apps Using Multimodal AI

AutoConsis, a joint effort by Meituan’s Quality Engineering team and Fudan University, automatically detects UI inconsistencies in mobile apps through a three‑stage pipeline—multimodal CLIP‑based region recognition, OCR‑enhanced information extraction guided by Chain‑of‑Thought prompts to a large language model, and rule‑based plus LLM‑driven consistency checks—enabling large‑scale deployment across 700 cities and thousands of pages to uncover pricing and semantic errors that manual testing misses, with future plans to add iterative agent feedback.

Mobile AppsUI Testingconsistency detection

0 likes · 14 min read

AutoConsis: Automated UI Consistency Detection for Mobile Apps Using Multimodal AI

Meituan Technology Team

Oct 31, 2024 · Artificial Intelligence

Selected Meituan Papers from CIKM 2024: Summaries of Eight Research Works

This article highlights eight Meituan research papers accepted at CIKM 2024—spanning self‑supervised sequential recommendation, rating‑consistent explanation generation, CTR prediction via recommendation pre‑training, cross‑domain interest transfer, multimodal vector retrieval, design‑aware poster layout, order‑fulfillment cycle‑time forecasting, and delivery‑scope substitution—offering insights from both internal and university collaborations.

AI researchCTR predictionCross‑Domain Recommendation

0 likes · 16 min read

Selected Meituan Papers from CIKM 2024: Summaries of Eight Research Works