Research Scientist in LLM Foundation Models (reasoning, planning & agent)

Founded in 2012, ByteDance's mission is to inspire creativity and enrich life. With a suite of more than a dozen products, including TikTok, Helo, and Resso, as well as platforms specific to the China market, including Toutiao, Douyin, and Xigua, ByteDance has made it easier and more fun for people to connect with, consume, and create content. Why Join Us Creation is the core of ByteDance's purpose. Our products are built to help imaginations thrive. This is doubly true of the teams that make our innovations possible. Together, we inspire creativity and enrich life - a mission we aim towards achieving every day. To us, every challenge, no matter how ambiguous, is an opportunity; to learn, to innovate, and to grow as one team. Status quo? Never. Courage? Always. At ByteDance, we create together and grow together. That's how we drive impact - for ourselves, our company, and the users we serve. Join us. About The Team: Founded in 2023, ByteDance Doubao Team is dedicated to crafting the industry's most advanced LLMs. We aim to lead global research and foster both technological and social progress. With a long-term vision and a strong commitment to the AI field, the Team conducts research in a range of areas including natural language processing (NLP), computer vision (CV), and speech recognition and generation. It boasts a robust international presence with labs and research facilities in China, Singapore, and the US. Leveraging substantial data and computing resources and through continued investment in these domains, our team has built a proprietary general-purpose model with multimodal capabilities. This model supports over 50 downstream business services including Doubao, Coze, and Dreamina, which are available to enterprise customers via the Volcano Engine. Currently, the Doubao App is the most used AIGC application in the China market. The vision of the team is to build a super intelligence system, accelerating society change in various areas such as natural sciences, economic production, and daily life, thereby enhancing the quality of life. The main areas of work include: 1. Reasoning and planning for foundation models. Enhance reasoning and planning throughout the entire development process, encompassing data acquisition, model evaluation,pretraining, SFT, reward modeling, and reinforcement learning, to bolster overall performance. 2. Synthesize large-scale, high-quality (multi-modal) data through methods such as rewriting, augmentation, and generation to improve the abilities of foundation models in various stages (pretraining, SFT, RLHF). 3. Solve complex tasks via system 2 thinking, leverage advanced decoding strategies such as MCTS, A*. 3. Investigate and implement robust evaluation methodologies to assess model performance at various stages, unravel the underlying mechanisms and sources of their abilities, and utilize this understanding to drive model improvements. 4. Teach foundation models to use tools, interact with APIs and code interpreters. Build agents and multi-agents to solve complex tasks.