Jianxin Ma
|
|
Biography
Currently, Jianxin Ma (马坚鑫) is doing research and engineering work within the Qwen Team, which was previously referred to as the M6 & OFA Team, at Alibaba Group. He earned his Master's degree in Computer Science and Technology in June 2020 from Tsinghua University, having also obtained his Bachelor's degree in the same field in June 2017 from the same university. During his postgraduate studies, he conducted research under the supervision of Wenwu Zhu and Peng Cui, focusing on the development of generative, interpretable, and controllable AI for structured data.
Research Interests
Building AI systems capable of interacting with the complex world (e.g., structured and multimodal signals) and learning from the collected experiences (e.g., memory, experiential learning, and automated training data collection).
Selected Publications (See Google Scholar for Full Publications)
- Qwen2 Technical Report. 2024. [model]
- Qwen Technical Report. 2023. [model]
- Pretrained Diffusion Models for Unified Human Motion Synthesis. 2022.
- OFA: Unifying Architectures, Tasks, and Modalities through a Simple Sequence-to-Sequence Learning Framework. ICML 2022. [code]
- M6-Rec: Generative Pretrained Language Models are Open-Ended Recommender Systems. 2022.
- M6-UFC: Unifying Multi-Modal Controls for Conditional Image Synthesis via Non-Autoregressive Generative Transformers. NeurIPS 2021.
- M6: A Chinese Multimodal Pretrainer. 2021.
- Contrastive Learning for Debiased Candidate Generation in Large-Scale Recommender Systems. KDD 2021. [code]
- Disentangled Self-Supervision in Sequential Recommenders. KDD 2020.
- Learning Disentangled Representations for Recommendation. NeurIPS 2019. [poster] [code & dataset]
- Disentangled Graph Convolutional Networks. ICML 2019. [code]
- Hierarchical Taxonomy Aware Network Embedding. KDD 2018. [code]
- DepthLGP: Learning Embeddings of Out-of-Sample Nodes in Dynamic Networks. AAAI 2018. [code][supp]