I am planning to apply for a young research group leader grant. (Specifically: the DFG Emmy Noether program.) This is for a six-year funding period, including funding for myself as a group leader and one or several Ph.D. students. The key part of the application is, of course, the research proposal, which is limited to 17 pages in length including background, objective, research plan and methods, and references.

Which brings me to my question: how on earth do you write a research plan for six years in a rapidly evolving theoretical field that is both interestingly ambitious and sufficiently concrete and grounded in reality?

Some more (possibly unnecessary) details:

I work in (a subfield of) machine learning, which may be a little bit of a special case due to the current speed of developments. I am quite certain that nobody could have usefully predicted what would be the topics and methods of current research interest six years ago. Heck, I'd like to claim that nobody could have predicted reliably what would be useful for a grad student to work on right now even 2 years ago.

Now, of course, I have some ideas for what would be useful and interesting to work on over maybe the next 2 years. With some stretch, I could probably work out an attractive, concrete, and reasonably grounded-in-reality proposal for the next three years and 2-3 full-time researchers. But six years? How do people deal with this? Are there useful strategies to follow? I could easily write down a long list of problems of interest that I'd like to work on, but there is no way I can foresee what will turn out to be the most useful methods to use for solving those? I assume reviewers will want to see a bit more than "this is an important problem that I plan to spend the last three years of the grant working on"... and of course, for a grant that covers more than a dozen full-time researcher-years, you also have to promise quite a lot.

  • 1
    As a fellow ML scientist, I would point out that there are many topics that are open problems for a long time, and still are even in the time of LLMs, like uncertainty, out of distribution detection/generalization, ethics, model biases, incorrect predictions, generalization, etc. These have been open problems for decades, so if you propose something in these lines, it easy to plan for 6 years.
    – Dr. Snoopy
    Commented Aug 17, 2023 at 17:49
  • @Snoopy Thanks for the comment, that is certainly true! My plans were also going in that direction, but still... I worked a bit on e.g. ood detection, and even there it seems like the field is developing quickly enough that (while certainly not solved!) any sufficiently specific plans you made 3 years ago are likely moot by now. Wouldn't you agree? I mean, there are also thousands of other ML researchers working on the exact same topic at the same time; you would hope that they would solve some problems...
    – Eike P.
    Commented Aug 17, 2023 at 17:56
  • Like I wrote, I find it easy to come up with interesting open problems to work on, but detailing specifically how I'm going to work on them over such a long time frame seems wild to me.
    – Eike P.
    Commented Aug 17, 2023 at 17:57

1 Answer 1


I've nothing specifically ML to offer. As a general strategy for a six year proposal I'd recommend writing about several current ideas you've already started on, with hopes/plans to have publishable work in the next year or two(*), then thoughts/speculations on how you might continue depending on how those results turn out. I doubt that the grantors actually expect a detailed six year plan.

(*) You might even propose to do work you have already (mostly) done but not yet published. That would guarantee a good interim report to the granting authority.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .