Liang-Yuan (Leo) Wu (@binomial14wu) profile picture

CS PhDing @NYUniversity | Taiwan🇹🇼 | Edinburgh 🏴󠁧󠁢󠁳󠁣󠁴󠁿 | Michigan 🇺🇸 | NYC 🗽

New York, USA
Joined June 2021
Liang-Yuan (Leo) Wu retweeted
A potentially controversial thought: We should make AI PhD coursework only optional. So many of our PhDs are tied in numerous classes, but: - Self-learning is easier than ever. - Many incoming PhDs are already strong researchers. - We need to free Profs' time for more research mentorship. - Urgent AI problems need free-thinking minds (like our academic students!), independent of industry for-profit directions. Courses are valuable. But they don't need to be mandatory, especially when so many problems with AI are threatening our society, and urgently need to be solved.
34
22
4
318
33,717
Liang-Yuan (Leo) Wu retweeted
How do you tell if authors fabricated their research results? Sometimes: they just tell you. Read, e.g., this caption carefully. Another banger of a post by my friend Nihar B. Shah on his blog, where he found at least 20+ such #ICLR2027 submissions: researchonresearch.blog/2026…
38
121
23
1,433
133,073
Liang-Yuan (Leo) Wu retweeted
An Anthropic researcher just told a room full of PhD students that their career plan no longer exists His 3 pieces of advice for students: 1. Use AI every day. If you're on the exponential, you'll never fall behind it. 2. Choose projects that get better when compute scales. The rest will be solved before you finish them. 3. Stop planning your career as if it were stable. The safe option no longer buys safety; it no longer exists. And a phrase that sums it all up: "Writing is thinking. If you outsource your writing, you outsource your thinking." The talk is titled "The End of the Anthropocene"
39
196
31
1,628
231,736
Liang-Yuan (Leo) Wu retweeted
The five stages of academic writing: 1. Procrastination 2. Panic 3. Caffeine overdose 4. Submitting a barely coherent mess 5. Swearing to do it differently next time
2
130
16
492
22,133
Liang-Yuan (Leo) Wu retweeted
New in JMIR Rehab: Assessing the Role of Medical Caption Technology to Support Physician-Patient Communication for Patients With Hearing Loss: Mixed Methods Pilot #Study dlvr.it/TQMsRF
2
3
87
Liang-Yuan (Leo) Wu retweeted
Okay so, we just found that over 50 papers published at @Neurips 2025 have AI hallucinations I don't think people realize how bad the slop is right now It's not just that researchers from @GoogleDeepMind, @Meta, @MIT, @Cambridge_Uni are using AI - they allowed LLMs to generate hallucinations in their papers and didn't notice at all. It's insane that these made it through peer review👇
277
1,368
415
6,205
1,007,606
Liang-Yuan (Leo) Wu retweeted
I'm 54. I wish someone had told me this at 25. 21 sentences that'll get you ahead of 95% of people:
46
389
9
3,311
1,679,109
Liang-Yuan (Leo) Wu retweeted
Exciting month ahead for the Soundability Lab! 11 paper, poster, and demo presentations, most at #ASSETS25 next week in Denver. Congrats to our team & collaborators — and special kudos to @binomial14wu for the Best Paper Nominee! 🔗 soundability.eecs.umich.edu
1
3
7
700
Liang-Yuan (Leo) Wu retweeted
In #ASSETS25, I will present a Demo of RAVEN 🐦‍⬛: Realtime Accessibility in Virtual ENvironments for Blind and Low-Vision People! Shoutout to the team @dj_hci, @venkypotluri, @Phylliiiis and @Chenglin791364! Find us at Demo Session C to try it out! Check out this short demo vid😆:
5
2
9
1,292
Liang-Yuan (Leo) Wu retweeted
Looking forward to presenting our #CHI2025 work with @dj_hci , @jayhersk , @binomial14wu , and Dr. Cecily Morrison on designing intent-driven AI systems for supporting deaf and hard of hearing people's sensemaking of auditory environments! Session Link: programs.sigchi.org/chi/2025…
3
14
969
Liang-Yuan (Leo) Wu retweeted
🥳 Huge congrats for the #ASSETS2024 Best Poster Award to Liang-Yuan Wu @binomial14wu, Andrea Kleiver & Dhruv Jain @dj_hci! CARTGPT is a human-#AI approach to generating highly accurate real-time captioning using #LLMs & ASR to fix in-situ CART errors in edge situations.
3
5
690
Liang-Yuan (Leo) Wu retweeted
🥳 Huge congrats for the #ASSETS2024 🏆 Best Poster Award to Liang-Yuan Wu @binomial14wu, Andrea Kleiver & Dhruv Jain @dj_hci! CARTGPT is a human-#AI approach to generating highly accurate real-time captioning using LLMs & ASR to fix in-situ CART errors in edge situations
Congrats @binomial14wu for the #ASSETS2024 best poster award. 🎉
2
8
1,047
Liang-Yuan (Leo) Wu retweeted
Congrats @binomial14wu for the #ASSETS2024 best poster award. 🎉
2
4
1
27
2,364
Liang-Yuan (Leo) Wu retweeted
Finally, @binomial14wu will present a poster on CARTGPT (3:30-4:15pm), a human-AI approach to generating highly accurate real-time captioning. Our approach uses LLMs & ASR to fix in-situ CART errors in edge situations (e.g., high noise, fast speakers). dl.acm.org/doi/10.1145/36635…
1
1
5
297
Liang-Yuan (Leo) Wu retweeted
Unfortunately, I missed #ASSETS2024, but look out for Soundability Lab students @jzh2o, @xinyun_c, & @binomial14wu presenting two papers, a poster, & a demo tomorrow. The key session is "3B: Sounds Around Us," 9 – 10:15 am. Details in the thread:
Students from the Soundability Lab will present 7 papers, posters, & demos at the upcoming #UIST2024 & #ASSETS2024 conferences. 🙌🎉 Come chat with us! @xinyun_c, Alex, & I will attend UIST. @jzh2o, @binomial14wu, @xinyun_c, & I will attend ASSETS. accessibility.eecs.umich.edu…
1
3
15
1,975
Liang-Yuan (Leo) Wu retweeted
Students from the Soundability Lab will present 7 papers, posters, & demos at the upcoming #UIST2024 & #ASSETS2024 conferences. 🙌🎉 Come chat with us! @xinyun_c, Alex, & I will attend UIST. @jzh2o, @binomial14wu, @xinyun_c, & I will attend ASSETS. accessibility.eecs.umich.edu…
1
5
1
29
4,562
Liang-Yuan (Leo) Wu retweeted
I hope everyone in Taiwan 🇹🇼 is okay after the earthquake. My thoughts are with everyone affected. ❤️
39
123
4
1,802
189,080
Liang-Yuan (Leo) Wu retweeted
The Happiness Equation and How AI Solves For It We are all engaged in pursuing and attaining happiness. Our Founding Fathers went so far as to call it out in the Declaration of Independence. So, can happiness be engineered? Short answer: Yes. The happiness equations are a good starting point Jonathan Haidt, a social psychologist, proposed an equation to describe happiness in his book "The Happiness Hypothesis." The equation is: H=S+C+V Where: - H is your overall happiness. - S is your biological set point, essentially the baseline level of happiness you're born with. - C represents the conditions of your life, such as your income, your occupation, and your marital status. - V stands for voluntary activities, or the things you choose to do, like exercising, pursuing hobbies, or spending time with loved ones. The idea behind this equation is to break down the complex concept of happiness into more manageable components: Set Point (S): This tends to be genetic and is hard to change. Some people are naturally cheerful while others tend to be depressed. Conditions (C): These are the external factors in your life. While you can change some (like getting a better job), others might be more fixed (like past life events). Voluntary Activities (V): These are perhaps the most controllable. They include the choices you make, the activities you engage in, and the way you see the world. Haidt's equation states that while you can't do much about your biological set point, you have some control over your life conditions and a lot of control over your voluntary activities. Therefore, if you want to increase your happiness (H), you should focus on the things you can change, particularly the V in the equation. AI can help directly solve for V by helping optimize voluntary activities like recommending great content on social media, personalizing entertainment options, and suggesting great dates on dating apps AI also started to contribute meaningfully to C (conditions in life). Using an AI assistant in your job can make your job more interesting. Let the AI do the heavy lifting, so you can focus on the creative and thinking parts. Another popular way to represent happiness is Happiness=w1​×Reality - w2​×(Expectations) This equation states that your level of happiness is determined by the gap between your expectations and your reality. If reality exceeds expectations, you're happy. If reality falls short of expectations, you're unhappy. The researchers who developed the above equation call it a “computational model of momentary subjective well-being.” w1 and w2 are weights that may add up to 1 and may be subjective on a per-person basis. You can also apply the equations to multiple situations like a job, a vacation, or if you will win the lottery. AI can help improve the reality part of this equation by increasing our standard of living. Inventions like AlphaFold have helped us solve the protein folding problem and will help invent personalized medicine and preventive cures to keep us healthy and happy. Personalized recommendations, AI assistants, and AI companions can all increase our happiness. AI can forecast outcomes and its data-driven insights and probabilities serve as a great "reality check" and help set expectations. AI chatbots can manage your emotional expectations In summary, AI can act as a sort of "happiness optimizer," helping you tweak both your reality and your expectations to get them more in sync
28
158
37
674
282,669
Liang-Yuan (Leo) Wu retweeted
So I finally did what I promised 3mo ago – We wrote a reflection paper on using LLMs to replicate crowdsourcing pipelines, based on an assignment in our Human-Centered NLP course! arxiv.org/pdf/2307.10168.pdf 20+ students implemented 7 crowd. pipelines from prior research & find… 1/
3
53
4
308
39,114
Liang-Yuan (Leo) Wu retweeted
Can language models help us do better search?🤔 🎉In #ACL2023 findings, we present EAR pipeline: 🎲sample multiple queries from LM 🎯rescore to select best query 🔍BM25 search 💪boost OpenQA accuracy to beat DPR/GAR arxiv: arxiv.org/abs/2305.17080 code: github.com/voidism/EAR – at Cambridge, MA
2
18
2
78
7,939