John Chih Liu retweetledi
(11/n) If you're interested in RAG, multi-hop reasoning, or evaluation of LLMs — give our paper a read and check out CRUMQs!
📄 Paper: arxiv.org/abs/2510.11956
🔗 Code: github.com/pybeebee/CRUMQs
English
John Chih Liu
2.6K posts
John Chih Liu
@drjohncliu
founder @Intelluron | @Penn PhD, CFA | ex-Credit Suisse, Lehman, BNP, @VanderbiltU, Citi, Digital Reasoning (acquired) | @creativedlab, @taiwangoldcard alum
Taipei Katılım Ekim 2008
309 Takip Edilen534 Takipçiler
John Chih Liu retweetledi
🚀 RAG systems excel at answering questions—but what happens when the corpus has NO answer or complex multi-hop reasoning is required?
Moreover, how can we build benchmarks to stress-test RAG systems in such settings in a realistic way?
See our new preprint to find out! 🧵👇
English
Bravo Kaili!
Gabrielle Kaili-May Liu@pybeebee
🎉 Delighted to announce that MetaFaith has been accepted to #EMNLP2025 Main! In this work we systematically study how well LLMs can express their internal uncertainty in words, offering a metacognition-inspired way to improve this ability 🧠✨ Check out more details below!👇
Lietuvių
John Chih Liu retweetledi
I will be presenting our work 𝗠𝗗𝗖𝘂𝗿𝗲 at #ACL2025NLP in Vienna this week! 🇦🇹
Come by if you’re interested in multi-doc reasoning and/or scalable creation of high-quality post-training data!
📍 Poster Session 4 @ Hall 4/5
🗓️ Wed, July 30 | 11-12:30
🔗 aclanthology.org/2025.acl-long.…
Gabrielle Kaili-May Liu@pybeebee
🔥Thrilled to introduce MDCure: A Scalable Pipeline for Multi-Document Instruction-Following 🔥 How can we systematically and scalably improve LLMs' ability to handle complex multi-document tasks? Check out our new preprint to find out! Details in 🧵 (1/n):
English
John Chih Liu retweetledi
If you use "AI agents" (LLMs that call tools) you need to be aware of the Lethal Trifecta
Any time you combine access to private data with exposure to untrusted content and the ability to externally communicate an attacker can trick the system into stealing your data!
English
John Chih Liu retweetledi
(14/n) If you're interested in LLM trustworthiness, uncertainty quantification, human-AI collaboration, or even metacognition — give our paper a read and check out MetaFaith! We'd love feedback or questions.
📄 Paper: arxiv.org/abs/2505.24858
🔗 Github: github.com/yale-nlp/MetaF…
English
John Chih Liu retweetledi
🔥 Excited to share MetaFaith: Understanding and Improving Faithful Natural Language Uncertainty Expression in LLMs🔥
How can we make LLMs talk about uncertainty in a way that truly reflects what they internally "know"?
Check out our new preprint to find out!
Details in 🧵(1/n):
English
Brett’s tweets are a great way to ride the pocket of AI
Brett Adcock@adcock_brett
A ton of great progress in AI and Robotics this week. I summarized everything from Unitree, OpenAI, Mirror Me, Microsoft, Physical Intelligence, Luma, Sakana AI and more. Here's everything you need to know and how to make sense out of it:
English
Excited to share an intriguing paper on complex multi-document tasks by @pybeebee! 📄✨A must-read for advancing LLM capabilities #AI #LLMs #Research
Gabrielle Kaili-May Liu@pybeebee
🔥Thrilled to introduce MDCure: A Scalable Pipeline for Multi-Document Instruction-Following 🔥 How can we systematically and scalably improve LLMs' ability to handle complex multi-document tasks? Check out our new preprint to find out! Details in 🧵 (1/n):
English
Deeply humbled to share my tenure promotion. This feels surreal and would not have been possible without the incredible support from so many people. To my mentors, colleagues, collaborators, and students - thank you for believing in me, inspiring me, and being there every step of the way. Your unwavering support means the world to me, and I am truly grateful to each of you. ❤️
UMIACS@umiacs
Congratulations to our eight faculty members who are receiving academic promotions in recognition of their continued excellence in research, scholarship and service to the @UofMaryland campus community and beyond. Learn more: go.umd.edu/1p8w
English
John Chih Liu retweetledi
📢The problem in model alignment no one talks about — the need for preference data, which costs $$$ and time!
Enter Kahneman-Tversky Optimization (KTO), which matches or exceeds DPO without paired preferences.
And with it, the largest-ever suite of feedback-aligned LLMs. 🧵
English
An in-depth MLOps course on taking machine learning from the development phase to production environments buff.ly/439Ma2z
English
Excellent Santa Fe Institute talk by Josh Tenenbaum on the limitations of GPT-4 and better approaches based on probabilistic world models and human-intuitive AI buff.ly/437slJ8
English
John Chih Liu retweetledi
Only 17 days left in the @SecurityBSides @BSidesLV CFP-get your talk proposals in!
Open tracks:
Breaking Ground
Common Ground
Ground Floor
Ground Truth
Hire Ground
@iamthecavalry
PasswordsCon
Training Ground
Underground
What're all these tracks @Daemontamer? Glad you asked! 🧵
English
@savvyRL Won't scaling the review process be easier now with the general availability of LLMs?
English
We have only 55 tiny papers to meta-review! Can I recruit 11 nice people to do 5 each before Monday? 🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏
Rosanne Liu@savvyRL
We are very close to finalizing decisions for all 219 submissions. We'd appreciate a few *emergency meta-reviewers* over the weekend to help us push through the finish line! Your job will be aggregating reviews to make a recommendation. Please reach out! iclr.dei.2023@gmail.com
English
John Chih Liu retweetledi
I wanted to give a primer on software architecture at Harvard systems reading group, so we're studying @lichess, one of the largest chess websites… and I made a code scavenger hunt!
notes.ekzhang.com/events/hsrg/li…
English
John Chih Liu retweetledi
Stop what you’re doing and read this entire post on the GPT-4 code interpreter plugin. This is completely bonkers and is going to change everything. andrewmayneblog.wordpress.com/2023/03/23/cha…
English
John Chih Liu retweetledi
Congrats to everyone involved in one of the big announcements today! 🎉
To those who are not: remember that your value is NOT whether you find yourself in one of the few in-groups. A healthy scientific field is not supposed to make you feel left out. You are doing great 🌞
English