Question 1: What exactly do you need to submit for a thesis proposal?
Most schools require three things: the basis for topic selection (background + current research status + significance), research content and technical roadmap, and schedule. First get the template for your school and major, then break down the tasks backward from the template sections—this saves more time than writing by feel.
Question 2: You still can't settle on a topic—what's the root cause?
The problem is that the research question hasn't been written as a single sentence. My approach is to first write clearly "what problem I want to solve, what the research object is, and where the boundaries are," then hand this sentence to a tool to break down the literature search dimensions—it comes up with the search terms, and I choose the direction, instead of repeatedly piecing together keywords in the search box. I use MuchSci's paper search assistant to validate the direction: the literature comes from 180 million items in Crossref, supports exact or fuzzy search by title, author, and DOI, and can also filter by year and citation count. When a match is found, it provides the DOI and automatically fills in citation information such as authors, journal, year, volume, issue, and page numbers. When searches return nothing, at least you can tell whether it's a "blank direction" or "the keywords weren't chosen correctly."
Question 3: How do you write the current research status so it doesn't look copied?
The status quo section isn't about rearranging abstracts; it answers "How far have others gotten, where are they stuck, and therefore what do I need to do?" Before writing, I first build a literature ledger (number, author-year, research question, method, conclusion, limitations, whether cited, citation location), have the tool break down each paper to fill in the research question, method, conclusion, and limitations, then reorder by theme—the paragraph order emerges on its own.
Question 4: Can't draw the technical roadmap?
Usually, that's because the steps haven't been broken down finely enough. I first use the Office Assistant Agent to generate a draft proposal PPT (input the project title, research question, and initial ideas; output includes research background, current status, methods, and technical roadmap structure), then ask of each page, "What are the inputs and outputs of this step?" Missing steps immediately become apparent.
Question 5: What if I can't finish reading all the English literature?
Split it into two levels: read reviews in full to build understanding; read the original text of highly relevant original research papers. For translation, I use paragraph-level comparative reading (bidirectional Chinese-English academic translation, preserving formula and terminology consistency, supporting full PDFs); the translation only helps me understand, and I always check parameters, values, and chemical formulas against the original text. Keep a separate glossary of "Chinese translation + original English term"; do not use the Chinese translations for searches.
Question 6: What will my advisor ask?
Mostly "Why are you doing this?" and "How will you prove it?" Before the defense, present the outline out loud to the air: if you only talk through these points, does the main thread hold up? Wherever you can't explain it clearly is where the structure has a problem.
In using this graduate student AI assistant, I divide the work into three tasks—"search, organize, produce a draft"—and do the rest myself.
Three boundaries: hold the line on academic integrity yourself—opinions and conclusions must be your own; translations are not factual evidence; confirm your target institution's plagiarism-check and AI-use rules yourself. Costs are billed by usage—MuchSci uses credits; paper analysis and document generation deduct credits based on actual processing volume; free registration includes 3,960 credits as a trial allowance; uploaded papers and data are not used for model training.