Official Alimama tech channel, showcasing all of Alimama's technical innovations.
The Ali Mama External Investment & Brand Algorithm Series compiles recent KDD, NAACL, and SIGIR research on ad allocation, pacing, traffic estimation, and creative optimization, offering a comprehensive overview of cutting‑edge techniques that enhance advertising efficiency and brand performance across digital platforms.
PosterMaker leverages a ControlNet‑based TextRenderNet with character‑level visual features and a reward‑driven foreground‑extension detector to generate high‑quality product posters that accurately render Chinese text (over 90% sentence accuracy) while preserving product fidelity, and is already deployed in Alibaba’s AI creative tool.
The article details how Apache Paimon, integrated as an external table format in Alibaba’s Dolphin OLAP engine, achieves millisecond‑level query latency and up to 10k QPS through ORC push‑down, manifest conversion, caching, concurrency, and encoding optimizations, outperforming StarRocks and Hologres.
UQABench introduces the first benchmark for assessing high‑density user embeddings that serve as soft prompts in LLM‑driven recommendation, featuring a three‑stage pre‑train‑align‑evaluate pipeline, seven personalized QA tasks, and findings that transformer encoders, side‑information, simple linear adapters, and larger models markedly improve accuracy while cutting input tokens to about five percent.
Alibaba’s Taobao Group unveiled its AIGX technology suite, including the RecGPT recommendation model, the AIGB generative bidding system, and a new AI‑generated video engine, detailing open‑source benchmarks, NeurIPS workshop participation, and measurable ROI improvements for e‑commerce advertising.
In 2024 Alibaba Mama outlines how large‑language models transform search advertising through a three‑line scaling roadmap—explicit inductive‑bias design, implicit compute growth, and auxiliary CV/NLP advances—implemented via a pre‑train/post‑train/CTR paradigm and the LUM user‑behavior model, promising gains in relevance, recall, and real‑time serving while highlighting inference efficiency challenges.
Alibaba Mama’s advertising platform migrated from a monolithic Flink‑Kafka pipeline to a layered Paimon lakehouse, adding DWS upsert support and multi‑layer storage, which delivers minute‑level data freshness, cuts latency by 2.5 hours, reduces resource use over 40 %, halves development effort and achieves ≥99.9 % availability.
This article analyzes the challenges of Alibaba Mama's brand advertising data warehouse built on a Lambda architecture, introduces Apache Paimon lake storage and Dolphin OLAP engine as a unified lakehouse solution, details implementation steps, performance gains, and business benefits across multiple advertising scenarios.
HighService, Alibaba’s Pythonic AI service framework, accelerates large‑model inference and maximizes GPU utilization by separating CPU‑GPU processes, offering out‑of‑the‑box quantization, parallelism and caching, and dynamically reallocating idle GPUs across clusters through a master‑worker scheduler to keep online latency low while boosting offline throughput for diffusion and LLM workloads.
The paper introduces a model‑based reinforcement‑learning auto‑bidding framework that learns a neural‑network environment model from real logs, generates confidence‑aware virtual data fused with real data, and employs the COMBO+MICRO stabilizer and a Lagrange‑dual method for ROI‑constrained bidding, delivering up to 6.8 % higher consumption, 5 % GMV growth and 3.7 % ROI improvement on Alibaba’s platform.
