Individual Meeting Guideline

01 / Research Philosophy

Problem-driven research

这一部分实际上是在介绍整个做 research 的思考顺序。Motivation 和 Problem 是核心： 先想清楚这篇工作想做什么，再定义 project/paper 关注的 problem，后面的 method 和 experiments 都应该从这个核心自然推出。

Full view
Motivation

Motivation 是最核心的，它通常可以压缩为一句话，即这篇工作想要做什么。

Define Boundary. 但 motivation 往往很大，一个工作不可能完全解决所有问题。 因此需要从 motivation 中精准定义这个 project/paper 所关注的 problem。

Problem

关注我过往的论文以及大家的写作可以看到，一个 problem 基本上由 challenge 定义。 一般我们会有 2 个 challenge。1 个 challenge 太单薄，3 个 challenge 通常太多。 如下，很多时候真正重点解决的是一个 major challenge。另一个是 minor challenge。

Major Challenge
Minor Challenge
Method
Technical Contribution 1

主要 contribution 应该对应这个 major challenge。

Technical Contribution 2

辅助 contribution 应该对应这个 minor challenge。

有了 challenge 和对应的 technical contribution，就要 propose method 去解决。 Method 的设计应该和这些 contribution 一一对应，而不是为了堆模块而堆模块。

Experiments

实验的 design 一般按照 research question。RQ 基本是自己 claim 自己的 tech 可以解决这个 challenge，到底解决了没有。每个 RQ 对应一个实验验证的方面。

RQ1 Overall results
RQ2 Exploratory analysis
RQ3 Other explanation or ablation study
02 / Objective

The objective of individual meeting

The individual meeting is to have discussions on important tasks such as the challenges, the methodologies, the experiment designs, etc.

Based on the presented material such as problem formulation, solutions, and experiments, I will give the feedback to improve them.

It is not held to check the progress or performance of your model (though performance is important).

So when you make the preparation, please keep in mind what do you want to discuss in the incoming individual meeting.

03 / Preparation

Be prepared to our individual meeting

As mentioned, the individual meeting is held for deep discussions. Hence, you should make a good preparation such as making some slides or updating the Overleaf draft.

For problem discussion

  • The motivation of the project. It should be compressed into one simple sentence: what does this work aim to do?
  • The boundary of the project/paper. Please clarify what problem this work focuses on and what it does not try to solve.
  • The major and minor challenges that define the problem. Please explain which challenge is the major one and why it matters.

For methodology discussion

When discussing the methodology, please illustrate clearly in your slides/draft:

  • The technical details of proposed method. This should try to be written in mathematical equations in the slides or the Overleaf draft.
  • The motivation/benefits of deploying this technique. Some preliminary experiments/observations may also be needed to make it convincing.

For experiment discussion

When you present the experiments, please illustrate in your slides/draft:

  • The experimental settings.
  • The motivation of the experiments.
  • The observations from the experiments.

Also, please prepare the plan of works in the next week in our individual meeting.

04 / Draft Updating

Treat the draft as part of research

Updating the draft is not a final-stage polishing task. It is how we force ourselves to clarify the research logic.

Writing is not only the final step of a paper. Writing is part of the research process.

Problem formulation Writing it down forces you to define exactly what problem you are solving.
Methodology Formalizing the method exposes unclear assumptions and missing details.
Experiments Writing RQs and settings makes baselines, metrics, fairness, and analysis more precise.

What will be done?

Example: Complete the ablation study for the representation alignment module on two datasets.

Why is it important?

Example: This experiment verifies whether the module is necessary for the main performance gain.

What should be delivered?

Example: A table, a short observation paragraph, and the corresponding update in the Overleaf draft.