Release notes for ACA for Large Models
This topic describes the latest updates to the Alibaba Cloud Certified Associate (ACA) for Large Models certification.
2026
V2.4: Course structure upgrade
Date: May 2026
In early 2026, coding agents such as Qwen Code, OpenClaw, and QoderWork expanded from developer tools to general office applications, making AI mastery a core skill for all employees. This upgrade shifts the course's focus from learning to use large models to building a personal AI work system, emphasizing how to control, manage, and collaborate with agents. The new version restructures all 11 lessons along an "Understand → Communicate → Act → Master → Implement" learning path to help students evolve from AI users to designers of AI work systems.
Core changes
Area | Description |
Shift in course focus | Shifted the focus from basic large model applications to controlling, managing, and collaborating with agents, and restructured the 11 lessons accordingly. |
Upgraded hands-on experience | Introduced real-world work scenarios, including the use of a desktop agent, knowledge base, tools, MCP, skills, and a self-check loop. |
Improved implementation skills | Strengthened the focus on business objectives, evaluation mechanisms, and the RIDE methodology to help students convert AI capabilities into executable plans. |
Changes to course structure
Lesson | Updated topic | Description |
1 | Understanding large models | Updated the history of AI development, introduced the concept of agents, and aligned the content with the Alibaba Cloud product ecosystem. |
2 | Typical use cases for large models | Restructured into a three-tiered model of use cases: generation assistant, execution agent, and decision support. |
3 | Optimizing input to improve response quality | Helps students master methods for effective communication with large models. |
4 | Enabling large models to answer domain-specific questions | Explains how to build an enterprise knowledge base service and ensure its quality through business objectives and evaluation mechanisms. |
5 | Improving large model performance in vertical domains with fine-tuning | The fine-tuning chapter is retained, focusing on its use cases, basic process, and how to choose between it, prompts, and RAG. |
6 | Installing a desktop agent and completing your first task | Shifted the focus from a plugin-centric view to an introduction to desktop agents, lowering the barrier to entry for non-technical students. |
7 | Connecting an agent to a knowledge base and tools | Explains how to use a personal AI agent for Q&A and introduces the use of tools and the principles of MCP. |
8 | Turning your work methods into reusable skills | Added new content on agent skills, explaining how to capture proven work methods as reusable skills. |
9 | Adding a self-check loop to an agent | Expanded from agent task execution to include output verification, self-correction, and closed-loop improvement. |
10 | Security and compliance for large model applications | Focuses on security risks in agent scenarios and adds five key principles. |
11 | Course summary | Supplementary learning materials have been integrated into relevant lessons. The summary now focuses on a methodology for implementing AI in enterprises. |
V2.3: Enterprise AI transformation methodology
Date: March 2026
In response to student feedback about being unsure how to apply their skills after completing the course, this update adds an enterprise AI transformation framework to the course summary. It introduces a closed-loop methodology, from use case identification to implementation validation, to help students apply the learned tools and methods to create actionable AI application plans.
Updated content
AI transformation framework |
|
V2.2: RAG evaluation and improvement
Date: January 2026
The previous RAG section of the course focused mainly on the setup process but lacked detail on how to assess effectiveness. This update adds basic methods for evaluating RAG performance and a systematic approach to improvement. It aims to provide students with a complete "build-validate-optimize" understanding, enhancing their ability to independently diagnose and improve Q&A systems.
Updated content
RAG evaluation by experts | Root cause analysis for Q&A |
|
|
2025
V2.1.1: Fine-tuning chapter optimizations
Date: December 2025
The original course's comparison between RAG and fine-tuning understated the value of RAG, which could mislead students into prioritizing fine-tuning. This update rewrites the relevant comparison sections, clarifying that fine-tuning is better suited for specific scenarios, such as when the deployment environment cannot provide a retrieval service or when knowledge needs to be internalized within the model. This helps students make more informed technology choices.
V2.1: Multi-lesson content optimizations
Date: May 2025
This update optimizes the explanatory sequence, diagrams, and case studies in several lessons to help students better understand key concepts such as the working principles of large models, tool calling, RAG, agents, and fine-tuning. These changes improve the accuracy, coherence, and readability of the course content.
Key optimizations
Optimized diagrams related to the basic concepts, training process, and product ecosystem of large models.
Adjusted case studies and explanations in the chapters on prompts, RAG, tool calling, and agents.
Added explanations for key topics, including RAG evaluation, agent architecture, and fine-tuning methods.
Three stages of model training (Lesson 1) |
|
V2.0: Exam syllabus update
Date: February 28, 2025
The weighting of topics in the original exam syllabus was imbalanced, and there was no mock exam for students to practice. This update adjusts the topic weighting and releases a mock exam to help students prepare more effectively. Additionally, this update corrects readability and formatting issues in some course content.
Updated content
Adjusted the weighting of exam topics.
Released a mock exam.
Improved the readability and formatting of some content.
V1.0 exam topics | Weight | V2.0 exam topics | Weight |
Large model concepts | 12% | Large model concepts | 18% |
Prompt engineering | 14% | Prompt engineering | 20% |
Enhancing large model applications | 24% | Enhancing large model applications | 20% |
Large model fine-tuning | 12% | Large model fine-tuning | 12% |
Large model agents | 20% | Large model agents | 14% |
Large model security and compliance | 12% | Large model security and compliance | 10% |
Multimodal large models | 6% | Multimodal large models | 6% |
V1.5: Lesson 1 update and Lesson 4 optimization
Date: January 26, 2025
The capabilities of the Tongyi Qianwen model are continuously updated, and the automated task examples in the original course were not closely related to daily office work. To help students better understand the model's capabilities and master reusable data analysis methods, this update modifies Lesson 1, "Understanding large models," and Lesson 4, "Automating tasks with an API." The former updates the introduction to the Tongyi Qianwen model and adds use case-based examples, while the latter adds content on using Lingma for data analysis.
Qianwen use cases | Lingma data analysis |
|
|
2024
V1.0: Course launch
Date: June 6, 2024
The first version of the Alibaba Cloud Certified Associate (ACA) for Large Models certification course launched. The course is designed for non-technical personnel and beginners to large models. It is structured linearly by technology module and consists of 11 lessons covering large model concepts, use cases, API calls, prompts, plugins, RAG, fine-tuning, agents, security and compliance, multimodal extensions, and a course summary.
The hands-on environment is based on the Model Studio (Bailian) platform. The exam consists of 50 online questions (35 single-choice and 15 multiple-choice). The passing score is 80.