Old News Archive
This page contains archived news items that were previously on the main about page.
2025 News Archive
- Aug 2025 — We released AgentFly, an agent training framework. The project is led by PhD student Renxi Wang, and the team will continue improving it. We’d be happy to hear from anyone interested in collaborating or providing feedback.
- June 2025 — I joined MBZUAI-IFM, working on foundation models with a focus on reasoning model development.
- Feb 2025 — We discovered that LLMs don’t work well when there are conflicting instructions. See our paper for more details.
- Feb 2025 — Our work SCALAR is out! It’s an uncheatable long-context LLM evaluation framework - welcome to use it!
- Jan 2025 — We organized a tutorial on Safety Issues for Generative AI at COLING.
- Jan 2025 — We release Libra-Leaderboard, a leaderboard for LLM evaluation. Website, X, LinkedIn.
2024 News Archive
- Dec 2024 — Our Survey paper on AI safety was accepted by JAIR. Title: Against The Achilles’ Heel: A Survey on Red Teaming for Generative Models. Paper.
- Nov 2024 — We release Nanda LLM a 10B Hindi-English bilingual LLM adapted on top of Llama3-8B. Paper LinkedIn
- Oct 2024 — We (LibrAI) are participating as exhibitors at GITEX, Dubai.
- July 2024 — We (LibrAI) are participating as exhibitors at WAIC - World Artificial Intelligence Conference.
- May 2024 — I have five co-authored papers accepted by ACL 2024, see you in Bangkok, Thailand!
- Apr 2024 — We did a comprehensive survey on generative AI attack and defense (red-teaming). We target a better and safer AI development. [Github][Paper]
- Apr 2024 — Our work Do-Not-Answer is featured in the latest Stanford AI Index Report 2024.
- Feb 2024 — We release Loki: An Open-source Tool for Fact Verification. We also release a commercial version with fancy UI and better user experience. TRY IT NOW!
- Dec 2023 — We release LLM360, A framework for open-source LLMs to foster transparency, trust, and collaborative research.
2023 News Archive
- Aug 2023 — We release Jais and Jais-chat: Arabic-centric foundation and instruction-tuned open generative large language models. Link to Technique Report, Blog.
- June 2023 — We release CMMLU, a massive multitask language understanding benchmark in Chinese.
- June 2023 — My Ph.D Thesis: Incorporating structured and unstructured information for geospatial question answering is now available online.
- May 2023 — We propose Bactrian-X, multilingual replicable instruction-following dataset and models.
- Feb 2023 — I moved to Abu Dhabi, UAE; and started my postdoctoral work at MBZUAI.