🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Greedy Coordinate Gradient is a useful method but takes a lot of time to run. We accelerated it by 5.6x using a method called probe sampling. The key idea behind probe sampling is to use a smaller draft model to filter unpromising candidates in the search. But the difficulty

Introducing our exciting new research on interpreting how large language models process multilingual input! #Multilingual #LLM #NLP

Introducing our new work on accelerating the GCG algorithm for attacking LLMs! #jailbreak #alignment #LLMs

🚀 We released SeaLLM v2.5, which obtains amazing performance on many Southeast Asia language benchmarks. 👇 Check out more about its highlights

Thrilled to share our new SeaLLM3 model! It is my honor to participate in the training of the base model!

🚀 Big Update Alert! 🌏 We're thrilled to announce a significant update on the SeaExam leaderboard, showcasing the best LLMs for Southeast Asian languages! Newly added: ✨ LLMs specialized for SEA languages: SeaLLMs-v3, SEA-LIONv2 🌐 Open LLMs emphasizing on multilingual

🧠 Sharing some observations from our very recent studies on LLM pruning, as I feel many phenomena already evolved: * Newer smaller models (~8B) are really tough to prune: e.g., pruning Llama-3 is much much more challenging than Llama-2 (because they are much more sophisticated

Did you know a tiny typo can throw off an LLM’s reasoning?🤯Our research shows how even small errors can disrupt AI's step-by-step thinking! We’re exploring what this means for LLM robustness and handling real-world, messy input. #AI aclanthology.org/2024.emnlp-mai… 1/3

Alibaba just released Babel on Hugging Face Open Multilingual Large Language Models Serving Over 90% of Global Speakers introduce two variants: Babel-9B, designed for efficient single-GPU inference and fine-tuning, and Babel-83B, which sets a new standard for open multilingual

Thanks Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) for promoting our new model! Yes, we use the model extension method to elevate Babel's performance ceiling, and the results show that it works well.

Babel🗼A multilingual LLM supporting 25 languages, released by the Alibaba DAMO team. Model: huggingface.co/collections/To… Paper: huggingface.co/papers/2503.00… ✨ 9B/83B chat & base ✨ Supports 25 languages: English, Chinese, Hindi, Spanish, Arabic, French, Bengali, Portuguese, Russian,

papers with code, models, datasets, apps, project pages and more now directly available on Hugging Face

🚀RL algorithms are shaping the post-training of LLMs, but how do their objectives connect? In this blog, I explore their relationships and provide a unified perspective through the Policy Gradient Theorem—the backbone of policy gradient methods. Dive in: lancelqf.github.io/note/llm_post_…

I have an opening for fully-funded 6-month visiting PhD student at SMU. Time: August 2025 - March 2026 Eligibility: PhD students from universities in Europe, North/South America, South-East Asia. Topic: NLP/LLM Email me for more details if you are interested~

🙋🏻‍♂️ I'm actively recruiting for multiple fully-funded positions in my group at SUTD! Topic: NLP / LLM / Multimodal Openings: 1 x Postdoc: start date is flexible 1 x PhD: earliest batch is Spring 2026 2-3 x Visiting students: fully funded for 6-12 months. Open to Bachelor /