ZHZisZZ / weak-to-strong-search

Weak-to-Strong Search: Align Large Language Models via Searching over Small Language Models
https://arxiv.org/abs/2405.19262
9 stars 1 forks source link

stop cheating #1

Open tpoisonooo opened 1 month ago

tpoisonooo commented 1 month ago

If you are not ready to open source your code, please do not create it .

arXiv is facing many researchers. I understand you may need claiming your work or whatever.

Posting it on arXiv should be sufficient for claiming your work, and you can always update the link afterwards. The paper claims to have code but says it will be available later (TBA), which makes readers feel deceived and will leave a bad impression on the corresponding and affiliated institutions, AKA shanghai ailab.

Perhaps many people in academia would think: 'Isn't arXiv just like that, where you can just post anything casually?' It's not only people in academia who use arXiv; those from second-tier universities also use it.

As a middle-aged person, I feel quite inappropriate to judge other people's work, and at the same time, I find it quite meaningless to face these academic games.

Now that the author stick to keep old title, then keep it.

tpoisonooo commented 1 month ago

you said your source code is here and leave a blank.

image

tpoisonooo commented 1 month ago

image

ZHZisZZ commented 1 month ago

I am just wrapping up some other camera-ready papers. Our code will be open-sourced later, for your enlightenment. By the way, could you kindly stop being rude, ignorant, and stuuuuuuuuuuuuuuuuuuuuupid?

ZHZisZZ commented 1 month ago

Why do you believe using placeholder link constitutes cheating? A public GitHub repository is an excellent platform for gathering constructive feedback from the public, rather than receiving ungrounded claims from you (i.e., cheating). This feedback helps us improve our work before its official release. TBA clearly conveys that we will upload our code as soon as possible (after I wrap up some camera ready papers). Using a placeholder repository is a common practice; for instance, the Alignment Handbook repository serves as a placeholder for several months until their codes are finalized. If I have violated the standards of Shanghai AI Lab, I am willing to accept your criticism. Otherwise, you should be accountable for your statements as a self-proclaimed representative of Shanghai AI Lab.

By the way, stop editing your messages significantly as if it is me who started this argument.

tpoisonooo commented 1 month ago

Why do you believe using placeholder link constitutes cheating?

Yes, github placeholder is an excellent platform for gathering feedback about paper, I never think about it before. My main concern is whether the implementation exists or not.

“stop cheating” is indeed rude expression.

self-proclaimed representative

I have reviewed my word, it seems there has been no indication of claiming to be a representative of anything.

I simply stated that I am also a reader, and when I see code, I will check which organization it belongs to, which is a very normal behavior.

Everyone has the day they leave here, it's a job, no special meaning.

tpoisonooo commented 1 month ago

For gathering feedback, we usually create a group and immediately send messages after open-sourcing the code. It's faster than GitHub.

tpoisonooo commented 1 month ago

Using a placeholder repository is a common practice

Yes it is a common practice, common practice != rightness. Each team has their own principle, so I said, I feel it's quite inappropriate to judge other people's work.

tpoisonooo commented 1 month ago

Why do you believe using placeholder link constitutes cheating?

It seems the core disagreement.

tpoisonooo commented 1 month ago

as if it is me who started this argument.

No. I start this argument.

ZHZisZZ commented 1 month ago

Please don't try to tell me what to do if you're not in a position to set the rules for the whole lab. If you knew my affiliations, you should have conveyed your opinions in a private channel. But since you've brought it up publicly, so be it.

tpoisonooo commented 1 month ago

Please don't try to tell me what to do

Since I not in a position to set rules, so I closed the issue before, you reopen it.

tpoisonooo commented 1 month ago

Can we close it now ?