Hunyuan Multimodal Reinforcement Learning Research Intern
Description
This role involves conducting research on Reinforcement Learning (RL) algorithms for multimodal models, including diffusion models for image, video, and 3D generation, and autoregressive models for multimodal understanding. The intern will also design and develop RL infrastructure and reward modeling strategies to improve training efficiency and stability, and explore next-generation RL paradigms. The position is for a PhD student in Computer Science or a related field with strong research capabilities and programming skills.
Responsibilities
- Conduct research on RL algorithms for multimodal models, including diffusion models for image, video, and 3D generation, autoregressive models for multimodal understanding, and potentially unified multimodal frameworks.
- Design and develop RL infrastructure and reward modeling strategies to enable efficient large-scale training, improve training stability, and mitigate reward hacking and related failure modes.
- Explore next-generation RL paradigms that more directly and effectively learn from environment feedback.
Requirements
- Currently enrolled as a PhD student in Computer Science or a closely related field.
- Demonstrated strong research capability, with publications in top-tier conferences such as ICML, NeurIPS, ICLR, CVPR, ICCV, ECCV, SIGGRAPH.
- Strong hands-on programming skills, with solid experience in deep learning system implementation, model training and inference optimization, CPU/GPU acceleration, and distributed training and inference.
- Prior experience with diffusion models, autoregressive models, and/or text-to-image or text-to-video generation is highly preferred.
- Participation in ACM/NOIP is a strong plus.
Benefits
- Eligible for 1 hour of paid sick leave for every 30 hours worked.
- Up to 13 paid holidays throughout the calendar year.
- Full-time interns are eligible to enroll in the Company-sponsored medical plan.
