A Data-Efficient Path to Multilingual LLMs: Language Expansion via Post-training PARAM$Δ$ Integration into Upcycled MoE May 8, 2026· Hao Zhou , Tianhao Li , Zhijun Wang , Shuaijie She , Linjuan Wu , Hao-Ran Wei , Baosong Yang , Jiajun Chen Shujian Huang · 0 min read Cite URL Type Conference paper Publication Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics Last updated on May 8, 2026 AEA: Adaptive Expert Allocation Improves Sentence Embeddings from Mixture-of-Experts LLM May 8, 2026 →