Business Unit

What the Role Entails

Research directions include but are not limited to:

- RL Algorithms for Reasoning Models：Design robust RL training recipes (PPO / GRPO / GSPO variants) for large-scale reasoning models. Tackle training instability, reward hacking, and policy collapse in long-horizon and async settings. Explore how to bridge the gap between RL post-training and genuine reasoning capability improvement.

- RL for Autonomous Agents: Build RL pipelines for long-horizon terminal agents and tool-use agents. Investigate credit assignment, exploration strategies, and self-evolving agent behaviors in complex interactive environments.

-Reward Modeling & Optimization: Develop reward signals and regularization techniques that go beyond outcome-based rewards. Explore token-level reward shaping, entropy-based regularization, and learned reward models that generalize across tasks.

Who We Look For

- Enrolled in a PhD or Master's program in computer science, machine learning or a related field. 

- Solid understanding of RL fundamentals (policy gradients, PPO, GRPO, etc.) and hands-on experience applying them to LLM training.

- Strong programming skills in Python and PyTorch; experience training models on multi-GPU setups, comfortable debugging training instability at scale.

- Ability to independently read, critique, and build on recent research papers.

- Published or submitted first-author papers at top ML/NLP venues (NeurIPS, ICML, ICLR, ACL, AAAI, EMNLP, etc.), or demonstrated equivalent research maturity through preprints and technical reports.

- Familiarity with LLM post-training (RLHF, DPO, GRPO), model merging, or agent frameworks (ReAct, tool-use) is a strong plus.

- Experience with large-scale distributed training (DeepSpeed, FSDP, Megatron) and open-source contributions are welcome.

Equal Employment Opportunity at Tencent

As an equal opportunity employer, we firmly believe that diverse voices fuel our innovation and allow us to better serve our users and the community. We foster an environment where every employee of Tencent feels supported and inspired to achieve individual and common goals.