Should scientific research have very watery ideas to be published?

【Written in front】

That is, I can only publish ideas in EI, Shuihui, and OA SCI journals. I think it is rubbish when I see it. Is it worth publishing?

Source: https://www.zhihu.com/question/372648294

Author: Xiaohou Feideuterium

I think many people have a misunderstanding that the quality of scientific research mainly depends on the quality of the original idea.

In fact, idea is important, but it is usually not thought of in a flash, but it is patched out after countless painful trial and error and thinking corrections. When a work is finally published, most of the original idea has been changed so that even its mother does not recognize it.

Therefore, what determines whether a paper can be published in the end is often not the quality of the initial idea, but your efforts to continuously correct the idea during the trial and error process.

Let’s show up and talk about it yourself. The mental journey of my doctoral project is probably like this:

Boss: Look at the XXX pancake, ah bah, that XXX has a good direction, and other similar articles have been published in sub-journals (it was published in 2004, and there have been no good results in the past ten years...)
Researching literature, it seems that I want to write simulation software by myself, so I spent half a year coding the code, and used it to write a small paper
After pulling it out, I found that the ceiling was sealed, but the foundation was not laid properly, and there was no underlying physical model at all, so I shifted my focus and started pushing the model
Re-investigated the literature and found that the physical core is XXX. Hey, no one seems to have calculated this thing, I think it can be done
After a long time, I found that the structure is too complicated, and the amount of calculation required is an astronomical figure—fuck, it turned out to be a big pit, no wonder no one did it
After all, counting a few special cases, it seems that you can write a small paper? Regardless
When sorting out the paper, I found that, hey, this part of the data seems to be a bit regular? It seems to be able to give a general formula? ——I am simply a genius, I want to publish the top issue of the thesis of the water ball!!!
Guess the physical law may be XXX, redesign the model for calculation
Damn, the conjecture and the result are completely different . It turns out that I am an idiot. Heroes, please start again
Sister, this conjecture has been falsified 50 years ago, leaving tears of ignorance, heroes, please start again
Damn chicken, there seems to be a bug in the program, this batch of data is for nothing, please try again
…
Constantly revise the thinking, after N times of iterations, finally change a universal physical model, the initial verification effect seems to be not bad?
Show the results to the boss, and they will criticize randomly: this place is wrong, that derivation is not rigorous, this picture needs to be filled with data
The boss said to aim high, dream big, add some macro simulations, and find a few experiments for comparison. The pictures should be more fancy, and the story should be more sexy——At this point, I have already laid down and let it go, what the hell?
Fighting wits and courage with the boss for N rounds, after making N patches, I finally have a systematic result, and I can contribute
rejected
Dissatisfied, I made up a dozen pictures, rebuttal
Reviewer A: This result is not bad, it can be published; Reviewer B: How dare you rebuttal even rubbish? then refuse
Edit: Hey, can you two unify the caliber, forget it, I will invite reviewer C to arbitrate
It may be because I helped the old lady cross the road yesterday and saved some character. Reviewer C actually praised me fiercely. After a few rounds of minor repairs, I finally published it.

Looking back, the original idea is like telling you that there is food in the forest in the north. This food may be the sour and astringent tree fruit on the edge of the deep forest, which can be easily picked by raising one's hand. But it may also be game in the depths of the forest, which requires a bloody battle through mountains and rivers, or even a detour from the south to reach it.

A picture is worth a thousand words version [1] :

Author: Assassinated

I want to share my views from the perspective of reviewers . In the past two years, the corresponding author has published more articles, and has been invited to review more and more manuscripts, so I have some thoughts on this issue. I am going to compare the two articles reviewed in 2019.

When I was a Ph.D. student, I often saw the first B of articles published by Daniel Labs abroad. Every time I want to yell, what the hell can I say? At that time, it was always blamed on "academic injustice". It is believed that resources are not equal, and foreign laboratories team up with each other, "it means the same thing as big V mutual praise and drainage". At that time, I thought, one day I will be a reviewer, and all the articles that are filled with water will be thrown into the toilet, and the first sentence of the review comments will be given to him "This is bullshit!".

But now views are slowly changing.

First buckle down the question. What does "water" mean?

The subject of the topic said that "scientific research has a lot of ideas". So first of all, there is an idea. No matter how much you dislike your own idea, it is also an idea. I personally think: there is no water idea, only water articles. These are two concepts. Let's talk about why slowly.

One of the two articles reviewed in 2019 was submitted to a journal under IEEE. The journal is a traditional good journal, and there are many very good articles. In the past two years, it has been squeezed by new journals, and its influence seems to have declined a bit, but the IF still has 1.3. The other is New Journal of Physics (IF3.8), which has a higher impact.

IEEE

IEEE comes from a certain university in Northwest China that I don't know. I was very happy to see that the article came from a domestic university. I know our field very well, so I know that this school does not have corresponding experimental conditions, so this article is theory plus numerical simulation . I like the fact that the article is typeset in LaTex. But after a careful reading, that’s not the case. I almost immediately wrote the review comments, and I had to reject it. The writing was too bad. If you want to reject after reading an abstract and introduction, how many slots do you have? English is not good, I, a Chinese who is used to Chinglish expressions, are confused. I guess it would be better to use Google Translate directly; the structure of the article is unreasonable, with one sentence and one sentence , and it feels very casual . Be sure to pay attention”; there are many graphs, but they are basically one curve and one graph. In fact, it is more intuitive to compare many curves together. Many heatmaps don't even have a colorbar.

All in all it's poorly written. But please note that I never said that this article is watery. Anyway, there is still something to read down. In fact, this thing is also improved on the basis of foreign laboratories. If you want to say plagiarism, no, there is something; if you want to say original, it is basically very trivial; and it is just a simulated article, it is hard to say whether it can be experimented with. In addition, I specifically searched for this team. The corresponding author of the article, Big Boss, has a good level and has experience studying abroad in Europe and the United States. I read his article specifically, and his English is absolutely no problem. Therefore, the boss of this article must have not read it. Therefore, in terms of writing and innovation, there is nothing wrong with rejecting this article.

NJP

NJP this is very interesting. I saw this article on ArXiv, and read it carefully because it was a bit related to my direction. The writing is very good, and the overall feeling is that it is a boutique. The idea is good, the article is well written, and the pictures are beautiful. And from the Daniel Lab at UCSD. The most important thing is that I have an article that can just cite this article as a theoretical support. The next day I received a review invitation from NJP. At that time, I thought that I had already read the article, and the review comments must be easy to write. In principle, it is recommended to publish after minor revisions.

Therefore, if there is no accident, IEEE rejects it, and NJP recommends it. However, the fate of these two articles in my hands was exactly the opposite.

IEEE This article is the first author of a doctoral student. So I think we should be careful. The first author is also a person who is about to graduate, but his first article is not seen in the publication list of the research group. I have a little bit of compassion. So desperately looking for highlights for this article. This is really difficult for reviewers. I must admit that the idea of this article is not very bright, and the writing is also poor. But if you can solve the technical problems "English, logic, pictures, etc.", this article may not be watery. So I wrote the review comments. I don't want to reject the article, and I don't want the author to feel that they wrote it so well that they can't revise it. So I wrote 28 comments. Except for technical ones, I clearly pointed out that this article is not well written, and the meaning of the work is not very meaningful. I do not recommend publishing it. At the same time, I sent a private message to the editor. I don't want to reject this article. As long as the author can solve my comments, I can still recommend it for publication in principle. So this article is now the end of the third round of review, and the article has indeed been improved a lot, and it is estimated that it will be published soon.

The NJP one was rejected. The corresponding author of the article has cooperated with our small boss for many years, and we often meet in meetings, and I also have an upcoming article that will be published at the same time as him. But I was still rejected. Because I read that article again. There is no problem on the surface, and it can stand any scrutiny. But the problem is with the references. Several of their own articles appear in the references. I read them all together. I also read the references of a certain article in the references. Then the problem comes. I originally thought the idea of this NJP article was very good. After reading their own articles, the idea has already been published by themselves, and it was published in the top physics journal PRL "I haven't read it before", so I thought the idea was very good at first. But now the idea is gone! If the innovation is gone, the article will be very watery. What's even more exasperating is that a key technology in this NJP article comes from the PRL "I found it from the references of the references" published by the co-author a few years ago, but it was not cited! Don't you quote your own work? I can only say that either the first author does not understand the research status in this field, or the co-author has not read this article. The first author is also a Ph.D. student, but I'm sorry, the principle issue should be rejected. Of course, I am not the only reviewer, another reviewer's opinion is also negative, so this beautiful article can only lie in ArXiv.

To sum up, it doesn't matter what ideas I think are watery, and they are all worth publishing. What if I can help others? The key is that the article cannot be watered. The meaning of the article is not water is that there must be your own dry goods in it. As for whether others think it is meaningful or not, let's not consider it first. Are periodicals and magazines divided into grades and grades? The story should be told well and logically. Be careful with the drawing. Change the English a few more times, "Because it's too bad for people to reject English."

In addition, whether you are in water or not will also score which stage you are in. It is good for graduate students to have ideas. What are you thinking? Send it if you can! Of course, if you are already famous, you should cherish your feathers. After all, it is not easy to establish reputation , so don't let yourself get screwed.

Author: Climber.pI

It's a matter of opinion. I only discuss purely theoretical subjects, not including experiments or even engineering. Brushing arXiv in the past two weeks has a little feeling.

At the beginning of last year (then) the mentor took the initiative to chat with me once, saying that you have to shoot high. For example, Muli Safra once told her at some point that the level of work a person does is roughly the level of his/her work during PhD. If you think about it for a while, you will find that this is indeed the case. The representative works of famous faculty usually have at least one piece of work during the PhD period or in the early post-doctoral period. For example, Muli Safra's representative work is the PCP theorem. One of the highest (paper) awards in theoretical computer science is the Godel Prize. Since the early 1990s, the results related to the PCP theorem have been won three times. And this result was his early post-doctoral work. Sanjeev Arora, who collaborated with him, was still a doctoral student at the time.

Then the tutor said, look at your previous result, it took you so long to write a thesis. Usually, if 90% of the time is spent doing research, then 90% of the time is spent writing papers. And if you want to write clearly, the time spent writing small results and high-impact results will not be much different. But a person's time is limited, and simply following up on the results of predecessors is often of little value. A digression is that my supervisor has several works that are posted on arXiv, but not published in official conferences/journals. It looks a bit clean freak.

In the end, the instructor said, you have to figure out which questions are meaningful and which ones are just trivial. Then you have to convince yourself that those meaningful problems are worth doing. The reason is not that the problem is famous or the technology you are solving is very interesting, but that you have your own reasons and judgments—even if many people don’t think the problem is Makes sense, but you should know what its meaning is. It’s not very meaningful to always do follow-up** (Supplement: middle-stage research sometimes looks like follow-up, but it’s not necessarily inferior to the so-called initial-stage research)**. Coincidentally, when I introduced my results to a teacher who won the best paper award at the top conference last year, he also said similar things. It is not enough to have new views on the existing results, but to have new ideas. .

If you continue to think about it, it is not difficult to realize that this is why publication is often a shallow standard (at least in my field):

Some people seem to have a lot of articles, and they are all small improvements to different problems, and there are no new ideas at all;
Some people don’t have many articles, but they are almost the pioneers (or early adopters) of the main technology of a certain problem. When this problem/technology is mentioned, this person will be one of the first names that everyone thinks of.

The former is common in those teams where half of the team works on a problem. The tutors have constant projects, and the students are happy to output papers steadily. However, it takes time to adapt to independent work after graduation (if you can adapt). The latter is commonly seen in PhDs who can survive in the top group in Europe or North America, and some of them will be later recognized rising stars . There are also some people who seem to have much higher quality and quantity of work during the post-doctoral period than during the PhD period, but if you look closely at their PhD thesis, you can often see the clues. Their later work also has a certain continuity with the early accumulation.

In my opinion, there are probably the following types of PhD students:

The best students are already collaborators with their tutors when they are studying for PhD, and they even find topics and bring their tutors to do research. For example, someone who won the best paper of F meeting last year; Put it politely in the acknowledgment.
Again, it should also be regarded as an advisor in the literal sense. As a top expert in the field, he can give you precise advice based on his own experience and overall view (including a deep understanding of different issues). Most of the people who do well are in the superposition state of this level and the previous level . In many cases, the choice of topics depends on the students themselves.
The next step is for the tutor to give the questions that have been sorted out clearly, and even have a general idea that is almost feasible, and then figure it out by himself. Most people who are doing well are the superposition of this layer and the previous layer, and the topic selection largely depends on the tutor.
No matter how bad it is, you can only do it by hand, including but not limited to the instructor himself, or the senior person in the group. The main reason why students can participate in this project is that other people are too lazy to write papers or do complicated calculations/proofs. But to put it bluntly, this is generally a routine for undergraduates to experience research.

So are there small results that should be written up and published? If the main problems you are working on cannot be advanced, it is normal to spend some time writing out the existing small results and try to post them in some places. It's just to figure out which is more important. Didn't Andrew Wiles accumulate a bunch of (under his standards) not so big results back then? Otherwise, it is very likely that they will not be able to survive-not everyone can spend six years on a problem, then retreat to the next best thing to adjust the direction slightly, and create a series of breakthroughs in one or two years.

【Project recommendation】

The core code library of top conference papers for Xiaobai: https://github.com/xmu-xiaoma666/External-Attention-pytorch

YOLO target detection library for Xiaobai: https://github.com/iscyy/yoloair

Analysis of papers for Xiaobai's top journal and conference: https://github.com/xmu-xiaoma666/FightingCV-Paper-Reading