Student Researcher [Seed LLM - Code Generation] - 2026 Start (PhD)

Description du poste:

About the team The Seed LLM Code Generation Team is dedicated to comprehensively enhancing the model's coding capabilities and building a bridge for AI to interact with the digital world. Currently, the team focuses on key technologies such as large-scale automated synthesis of coding competitions/engineering problems, reinforcement learning for code agents, and construction of high-quality code pre-training datasets. PhD internships at ByteDance provide students with the opportunity to actively contribute to our products and research, and to the organization's future plans and emerging technologies. Our dynamic internship experience blends hands-on learning, enriching community-building and development events, and collaboration with industry experts. Applications will be reviewed on a rolling basis - we encourage you to apply early. Please state your availability clearly in your resume (Start date, End date). Responsibilities - Develop methods for code generation and editing using large language models, including improving performance on tasks such as synthesis, repair, documentation, and test generation. - Conduct research on self-evolving agents that can learn to write, edit, and optimize code over time with minimal supervision. - Explore multi-agent reinforcement learning settings where agents collaborate, compete, or communicate to solve complex programming tasks. - Investigate instruction tuning, multi-turn prompting, retrieval-augmented generation, and RL for program synthesis. - Build benchmarks and tools for evaluating model performance in code understanding, collaborative reasoning, and long-horizon programming scenarios

Profil requis du candidat:

Minimum Qualifications: - Currently pursuing a PhD in Computer Science, Machine Learning, Programming Systems, or a related field. - Research experience in code generation, program synthesis, or LLMs for software engineering. - First-author publications in top venues such as NeurIPS, ICLR, ICML, PLDI, or OOPSLA. - Familiarity with LLM toolchains (e.g., HuggingFace, PyTorch) and structured code/data handling. Preferred Qualifications: - Experience with reinforcement learning or agent-based learning in structured environments. - Background in multi-agent coordination, curriculum learning, or self-improving systems. - Understanding of software reasoning tasks such as static analysis, refactoring, or automated debugging. - Familiarity with open code benchmarks (e.g., HumanEval, MBPP, CodeContests, SWE-bench)