Nhat Tan Bui

Research Associate @CMU_Robotics

[Nhat Tan Bui @nhattanbui22]

@WeiyanShi6 Your research is so interesting, hope we can collab in the future

[Minh-Quan Le @lmquancs]

1/n When we explain an idea at a whiteboard, we talk and draw at the same time. We notice what’s missing, change the sketch, and find a better way to explain it.

Our NeurIPS 2026 work, CO₂Jump, brings this back-and-forth to AI: understanding and generating images together. 🧠🎨

[Nhat Tan Bui @nhattanbui22]

@lmquancs Nice work, Quan!!

[Nhat Tan Bui @nhattanbui22]

@MangQiuyang @furongh They also focus on novelty, but novelty should be identified by problem analysis, method, experiments, insights; not purely by techniques that used to build the method.

Glad to see you point out this. They are multiple factors for good papers beside number and novelty.

[Nhat Tan Bui @nhattanbui22]

@thientrangngv great work, Trang!!! And I really like the way you write down your blog!

[Nhat Tan Bui @nhattanbui22]

@DianboLiu @andrewgwils Then, researchers from developing countries can no longer submit papers. If the paper gets accepted, authors already pay for conference fee. Research should be free for all researchers, not be charged like this.

[Nhat Tan Bui @nhattanbui22]

Such a great job!

[Trang @thientrangngv]

CliffCompaction, our autocompaction tool for coding agents, is out!

Sessions run for millions of tokens, agents stay on track, and costs drop by up to 50%.

Paper: https://t.co/OdPp6Mt1Gh
Code: https://t.co/39Mg0ZJZ9C
PyPI: https://t.co/yOSIP3KCUH
Blog: https://t.co/cZdN9ZbURf

[Nhat Tan Bui @nhattanbui22]

@j_dekoninck Suppose you work on theoretical stuff, I guarantee you’ll have a huge slop index.

[Nhat Tan Bui @nhattanbui22]

@trungdt880 Nice work! I wonder which GPUs and how many of them you need to train the student, anh?

[Nhat Tan Bui @nhattanbui22]

@tallinzen @AnkaReuel @Yminglai this is the best solution so far, and I'm happy that you are also a professor. Hope other profs/companies can take a look at other materials (like SoPs to understand candidate's mindsets).

If papers aren't the only metric, it would reduce a lot of cheating.

[Nhat Tan Bui @nhattanbui22]

@AnkaReuel Not a complete solution, hope profs/companies look beyond papers and consider other things (e.g. SoPs for candidates’ mindsets).

If papers aren't the main metric, people wouldn’t feel pressured to get them by any means.

[Nhat Tan Bui @nhattanbui22]

@AnkaReuel This pressure forces people at the bottom of the system to find ways to cheat just to survive (of course, I think that is wrong).

So professors, PIs, and companies are also part of this broken system. We should think about how to fix the problem at its root....

[Nhat Tan Bui @nhattanbui22]

@AnkaReuel Look at the current PhD admissions/job market. If you don't have top-tier papers, it can be extremely difficult to get a PhD position, and I believe very few professors would even agree to interview a student from a developing country with no publications.

[Nhat Tan Bui @nhattanbui22]

@SuJinyan6 Thanks for bravely sharing your story. Life is your own, so do what you love, stay away from negativity, and hopefully, we’ll succeed someday!!!

Also, another reason to do a PhD is simple: you enjoy research and get to choose what you want to work on.

[Nhat Tan Bui @nhattanbui22]

@JandialSurgan @AviralChharia Thanks, Surgan, for all your help!

[Nhat Tan Bui @nhattanbui22]

@thaoshibe $20 subscriptions should be ok chi 🤣

[Ankit Gupta @agupta]

8. @JandialSurgan discusses On the Fine-Grained Planning Abilities of VLM Web Agents

[Nhat Tan Bui @nhattanbui22]

@pbdang2000 Great work bro!!!

[Nhat Tan Bui @nhattanbui22]

Great example of cognitive-inspired computer vision!

[Thao Nguyen (Shibe) @thaoshibe]

x threads 🌎🍑

here: https://t.co/E27if5k70x

or here: https://t.co/Pg1wQioqr8

🤗