Multimodal Reinforcement Learning Algorithm Researcher

Roles &Responsibilities

1.Conduct research on reinforcement learning algorithms formultimodal models, including diffusion models for image and video generation,autoregressive models for multimodal understanding, and cutting-edge unifiedmultimodal frameworks.

2.Design and develop reinforcement learning trainingframeworks and reward modeling strategies to enable efficient large-scaletraining, improve training stability, and address issues such as rewardhacking.

3.Explore next-generation reinforcement learning paradigmsthat enable more direct and efficient learning from environmental feedback.

Skills Required

1.Bachelor's degree or above in Computer Science or relatedfields.

2.Excellent research capabilities with publications in topconferences including ICML, NeurIPS, ICLR, CVPR, ICCV, ECCV, SIGGRAPH, etc.

3.Strong engineering and programming skills, withexperience in deep learning system implementation, model training and inferenceoptimization, CPU/GPU acceleration, and distributed training and inference.

4.Preference given to candidates with experience indiffusion models, autoregressive models, text-to-image / text-to-videogeneration.

5.Preference given to candidates withparticipation experience in ACM/NOIP (Informatics Olympiad).