Official Didi technology account
The article details the creation of the didi-ride-skill for OpenClaw, explaining how a single voice command triggers a full ride‑hailing workflow, the underlying MCP toolset, engineering trade‑offs such as file splitting, attention handling, cron isolation, key management, testing strategies, and future roadmap.
The STAPO (Spurious‑Token‑Aware Policy Optimization) algorithm, introduced by Tsinghua University's iDLab and Didi's Deep Sea Lab, tackles policy‑entropy instability and performance oscillation in reinforcement‑learning fine‑tuning of large models by mathematically analyzing token collision probability, defining spurious tokens, and applying a Silencing Spurious Tokens mechanism that yields state‑of‑the‑art results on multiple math‑reasoning benchmarks.
Huatuo, the open‑source deep‑observability platform backed by Didi, now supports real‑time monitoring of MetaX GPUs, offering detailed hardware metrics via Docker or Kubernetes deployments and exposing them through a /metrics endpoint for cloud‑native AI and operations use cases.
On November 3, Didi Open‑Source presented its ecosystem and four flagship projects—XIAOJUSURVEY, HUATUO, MPX, and KnowStreaming—to over a hundred Zhejiang University software students, sharing insights on enterprise‑grade open‑source practices, cloud‑native observability, cross‑platform development, and the role of open source in talent cultivation.
This article explores how AI‑enhanced IDEs can transform front‑end development by streamlining requirement analysis, technical design, UI automation, code refactoring, test generation, and code review, offering concrete prompts, workflow tips, and best practices to dramatically improve efficiency and code quality.
Explore how the Bellman equation underpins reinforcement learning, comparing Dynamic Programming, Monte Carlo, and Temporal‑Difference methods, and discover why TD’s low‑variance, online updates make it a powerful bridge between model‑based planning and sample‑based learning.
This article deeply explores the ordered consumption mechanisms of RocketMQ and its derivative DDMQ, explaining the implementation differences, source‑code details of sequential sending, broker locking, client‑side processing, and compares the two systems' approaches to guarantee message order.
HUATUO, Didi's open‑source cloud‑native observability project, leverages BPF‑based low‑overhead kernel tracing, unified metric and event frameworks, automatic flame‑graph generation, and seamless integration with Prometheus, Grafana and Elasticsearch to provide panoramic, zero‑intrusive monitoring and continuous performance profiling for complex production environments.
This article presents a unified framework for efficiently estimating causal treatment effects on exponential‑family outcomes, extending target regularization beyond Gaussian assumptions, deriving bias analysis for plug‑in estimators, proposing DR and TMLE‑based estimators, and validating them on synthetic and real datasets.
This article presents a data‑fusion framework that combines long‑term observational data and short‑term randomized experiments to identify and estimate long‑term heterogeneous dose‑response curves under continuous treatments and unobserved confounders, using reweighting, optimal transport, and balanced representation learning.
