Abe Hou

@abe_hou

PhD student at @stanfordnlp @stanford. Undergrad at @jhuclsp.

Stanford, CA

abehou.com

Replies Reposts Media

Abe Hou
@abe_hou
Jul 31
@ben_vandurme and @DanielKhashabi are super amazing mentors. They are supportive, technical, and attentive - and everybody else at JHU is also very collaborative - strongly recommend❤️😁
Daniel Khashabi 🕊️
@DanielKhashabi
May 26
@ben_vandurme and I are recruiting multiple postdoc fellows at JHU. We're looking for candidates w/ strong record in language models, reasoning, coding agents, and/or AI for science. Interested candidates should send their CV and a brief summary of their research interests to
1
Abe Hou
@abe_hou
Jul 22
I actually looked into this a bit more and asked chatgpt its top 5 preferences: Japan, Italy, New Zealand, Mexico, Norway. I then asked, what are your LAST five? To my surprise, it actually answered: Kuwait, Bahrain, Luxemberg, Brunei, and Vatican City after pivoting to "travel
Superman
@thesupermannx
Jul 20
Researchers proved every major LLM is secretly obsessed with Japan. And they finally figured out why. For years, we’ve been told that AI is entirely Western-centric, that it just reflects Silicon Valley and American values. A landmark paper by Cardiff and Basque researchers
1
Abe Hou
@abe_hou
Jul 22
So many subtle ways mode collapse can happen, and I am glad there are more and more great lm diversity work such as Verbalized Sampling is paying attention to this issue!
Weiyan Shi
@shi_weiyan
Jul 22
"It emerges after pre-training, during SFT and alignment when humans teach the AI how to behave." In Verbalized Sampling, we also found that humans tend to prefer typical answers during alignment, and proved this preference is a root cause of mode collapse. You can bypass it
Abe Hou
@abe_hou
Jul 14
Super exciting and impactful project! Can't wait to check out🧐🤠
Dominik Stammbach
@dominsta_nlp
Jul 8
How can AI support public defense? We have worked on this question for the last two years, and have (1) built an AI retrieval tool for the NJ Office of the Public Defender and (2) conducted qualitative interviews with defenders. Our main insights 👇(1/12)
1
Abe Hou
@abe_hou
Jul 14
Exciting direction! Looking forward to seeing more cool applications and follow-up of this🔥🔥
Yijia Shao
@EchoShao8899
Jul 14
AI is getting incredibly powerful, yet most of that power only reaches software domain. The next frontier isn't building more agents, but a collaboration layer between humans and AI. As a first step, we're open-sourcing Coco: a proactive co-assistant that help at the right time.