Implement a project using SpringBoot 3.2.0 . With a GET method HelloWorld API that requires SpringSecurity authentication, which returns a string saying "Hello World". And a username-password login API (could use username test and password 123456). Submit a GitHub link.
笔试题-爬虫:
Using both Playwright and DrissionPage to crawl https://mitadmissions.org/blogs/, save and submit the results to a CSV file. Your headers should include the following:
Title | Author | Comment Count | Time | Article Content | Images In Article|
笔试题-Python:
Implement a demo of streaming API with fast API.
笔试题-算法:
Auto Labeling with LLM / VLM / Other foundation models:
Choose any public dataset (text, image, or multimodal) that you are familiar with, and demonstrate how a foundation model (e.g., LLM, VLM) can automatically generate labels for it.
What You Need to Do:
1.
Dataset and Task Description
◦
Which dataset did you choose?
◦
What is the input feature (e.g., text, image, etc.)?
◦
What is the labeling target (e.g., sentiment, topic, category, object type, etc.)?
2.
Model and Method
◦
Which model did you use (e.g., GPT, Qwen, CLIP, etc.)?
◦
How did you use it? (API or local inference)
◦
Describe your prompt or input format design.
3.
Runnable Code (Core Part)
◦
Provide runnable code (preferably Python) that performs labeling.
◦
Show how you read data, call the model, and output results.
◦
Include a few example outputs (sample + model label).
4.
Result Analysis
◦
Describe what patterns you observe in the model’s labeling results.
◦
When does the model perform well or fail?
◦
If human labels exist, briefly compare accuracy or agreement.
5.
Human vs Model Labeling
◦
Briefly compare their advantages and limitations, such as: speed, cost, consistency, semantic understanding, or bias.
6.
Improvement Ideas (Short Discussion)
◦
Propose at least two possible improvements, such as: better prompts, multi-model voting, confidence filtering, or human-in-the-loop refinement.
Submission:
Please submit a compressed folder including:
•
demo_code.py or a Jupyter Notebook;
•
Example output file;
•
A short written report (Markdown or PDF, 1–2 pages) describing your dataset, model, results, and insights.