Ziyi did a great job to understand to what extent LLMs can show social intelligence in a game other than just win rate. She has enthusiasm to study/interpret LLM behaviors. Check the paper for more interesting findings and we appreciate your feedback!

#naacl2024 led to my first time in CDMX and I loved it. Coorganizing NAACL SRW and trustnlp at #naacl2024 was a blast, was stoked for solid reception of my tokenization paper🥹, and incredibly grateful for the mountains just outside the city🌋🏍️. Thank you, Mexico City.

Awesome to see yet another successful completion of our trustnlp workshop. Looking forward to the next one.

🚨Looking for a fast and effective contradiction retrieval method? In our new paper, we propose a sparse-aware contrastive learning to beat traditional bi-encoder methods with 37.2% performance gain on average! 🚀Project: sparsecl.github.io 📰Paper: arxiv.org/pdf/2406.10746

How to efficiently detect and correct the unfaithful outputs from your retrieval-augmented language model (RALM)'s long-form outputs? Introducing SynCheck and FOD, a suite of synchronous model-centric faithfulness monitoring and intervention algorithms for RALMs.

Proposing Ctrl-G, a neurosymbolic framework that enables arbitrary LLMs to follow logical constraints (length control, infilling …) with 100% guarantees. Ctrl-G beats GPT4 on the task of text editing by >30% higher satisfaction rate in human eval. arxiv.org/abs/2406.13892

I'm excited to join Prof. James Zou group as a postdoc scholar, aiming to push the boundaries of AI for scientific discovery #AI4Science. I've had an incredible and rewarding time with the uclanlp group and the VCLA group UCLA Computer Science. Deeply grateful to all my mentors,

🚨WARNING: Instructing T2I models to depict “diverse” people will severely harm historical factuality!🤯In this work, we benchmark the evaluation of this “FACTUALITY TAX” of diversity intervention prompts, and propose Factuality-Augmented Intervention (FAI) to resolve the issue.

Our pick of the week by Andrea Piergentili: "The Tail Wagging the Dog: Dataset Construction Biases of Social Bias Benchmarks" by Nikil Selvam, Sunipa Dev, Daniel Khashabi 🕊️, tusharkhot, Kai-Wei Chang (ACL 2023). #bias #socialbias #benchmark #ACL2023NLP #NLP #NLProc

Looking forward to this new challenge organized by Amazon -- amazon.science/trusted-ai-cha…

🚨 Funding opportunity🚨 If your group is working on code generation, AI safety, Red Teaming, or other dim. of Responsible AI, consider participating in Amazon's Trusted AI Challenge on coding security. The challenge is structured as a tournament where you can join as a Red Team

New research shows how to supercharge LLMs with knowledge graphs. Our "Tree Of Traversals" algorithm lets LLMs reason over multiple KGs without retraining. This boosts performance on complex QA tasks. Paper: amazon.science/publications/t… Github repo: github.com/amazon-science…

Pleased to share that we have two papers presented and one co-organized workshop at #ICML2024! 🎉 Huge thanks to all collaborators! 🙌 SciBench: Poster: icml.cc/virtual/2024/p… Website: scibench-ucla.github.io SPHINX-X: Poster: icml.cc/virtual/2024/p… Paper:

I will be presenting Repoformer in Oral 4D at Hall A1, 5PM tomorrow 7/24. You're also welcome to stop by our poster at #700 poster session 4, tomorrow 1:30-3pm. Looking forward to chatting with you! #ICML2024

We are hosting wonderful NLP colleagues at KAIST on their way to ACL Bangkok! 🤩 On-site registration is closed, but the talks will be broadcast on Zoom. Please join us! Date/Time: Aug 10, 2024, 10:05-12:30 KST (UCT+9) Parallel Session 1: Advanced Language Models and AI

Excited to announce that Tree-of-Traversals will be presented at #ACL2024 Main Conference! Tree-of-Traversals combines 🧠 LLM Reasoning 🔎 Tree search over actions 🎬 📚 Knowledge graphs ✅ Easy to implement interface ✅ Multiple KGs together ✅ Black-box LLM Keep reading 🧵/11

Pre #acl2024 talks, posters, and food 🥘🍝🍲🍛🥙 We Sunipa Dev Kai-Wei Chang Tristan Naumann Isabelle Augenstein Scott Hale Jose Camacho-Collados Pang Wei Koh Minjoon Seo JinYeong Bak and Yoon Kim, (with Mohit Bansal and Violet Peng joining online) had a blast thanks to the amazing students