基本信息
导师姓名：秦涛
担任职务：北京中关村学院准聘副院长
主要研究领域：深度学习、强化学习、科学智能、大语言模型
简介：中国科学技术大学客座教授、博士生导师， ACM、IEEE 资深会员，研究成果被引用超过34,000次，h指数80+，i10指数250+。曾任微软全球研究合伙人，微软科学智能研究院亚洲区负责人。研究领域涵盖深度学习、强化学习以及它们在自然科学、自然语言处理、语音和图像处理等方面的应用。 近期的研究重点是AI与自然科学的交叉，旨在为药物研发、生命科学、材料设计等自然科学多个领域设计基座大模型和快速算法。
个人经历
教育经历：
2008年博士毕业于清华大学电子工程系，获工学博士学位
2003年本科毕业于清华大学电子工程系，获工学学士学位
工作经历：
2022-2025 微软科学智能研究院，全球研究合伙人
2008-2022 微软亚洲研究院资深首席研究员/经理
学术兼职：
中国科学技术大学兼职教授
科学研究
学术专著：
l Tao Qin. Dual Learning, Springer 2020.
代表性学术论文：
l NatureLM: Deciphering the Language of Nature for Scientific Discovery. arXiv 2025.
l TamGen: drug design with target-aware molecule generation through a chemical language model. Nature Communications 2024.
l HybriDNA: A Hybrid Transformer-Mamba2 Long-Range DNA Language Model. arXiv 2025.
l E2Former: A Linear-time Efficient and Equivariant Transformer for Scalable Molecular Modeling. arXiv 2025.
l Accelerating protein engineering with fitnesslandscape modeling and reinforcement learning. bioRxiv 2023.
l BioGPT: generative pre-trained transformer for biomedical text generation and mining. Briefings in Bioinformatics 2022
l The Impact of Large Language Models on Scientific Discovery: a Preliminary Study using GPT-4. arXiv 2023.
l FABind: Fast and Accurate Protein-Ligand Binding. NeurIPS 2023.
l SMT-DTA: Improving Drug-Target Affinity Prediction with Semi-supervised Multi-task Training. Briefings in Bioinformatics 2023.
l Pre-training Antibody Language Models for Antigen-Specific Computational Antibody Design. KDD 2023.
l Dual-view Molecular Pre-training. KDD 2023.
l Retrosynthetic Planning with Dual Value Networks. ICML 2023.
l De Novo Molecular Generation via Connection-aware Motif Mining. ICLR 2023.
l O-GNN: incorporating ring priors into molecular modeling. ICLR 2023.
l R2-DDI: Relation-aware Feature Refinement for Drug-Drug Interaction Prediction. Briefings in Bioinformatics 2022.
l Direct Molecular Conformation Generation. TMLR 2022.
l Naturalspeech 3: Zero-shot speech synthesis with factorized codec and diffusion models. arXiv preprint 2023
l Learning to rank: from pairwise approach to listwise approach. International Conference on Machine Learning (ICML) 2007
l Fastspeech 2: Fast and high-quality end-to-end text to speech. International Conference on Learning Representations (ICLR) 2021
l MPnet: Masked and permuted pre-training for language understanding. NeurIPS 2020
l Fastspeech: Fast. robust and controllable text to speech. NeurIPS 2019
l Generalizing to unseen domains: A survey on domain generalization. IEEE Transactions on Knowledge and Data Engineering (TKDE) 2022
l Mass: Masked sequence to sequence pre-training for language generation. International Conference on Machine Learning (ICML) 2019
l Dual learning for machine translation. NeurIPS 2016
l Neural architecture optimization. NeurIPS 2018
l Achieving human parity on automatic Chinese to English news translation. arXiv preprint 2018
l LETOR: A benchmark collection for research on learning to rank for information retrieval. Information Retrieval Journal 2010
l R-drop: Regularized dropout for neural networks. NeurIPS 2021
l Incorporating BERT into neural machine translation. ICLR 2020
l A survey on neural speech synthesis. arXiv preprint 2021
l Introducing LETOR 4.0 datasets. arXiv preprint 2013
l Can generalist foundation models outcompete special-purpose tuning? Case study in medicine. arXiv preprint 2023
l An empirical study on learning to rank of tweets. ACM SIGIR 2008
l Image-to-image translation: Methods and applications. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 2020
l Feature selection for ranking. European Conference on Machine Learning (ECML) 2003
lRepresentation degeneration problem in training natural language generation models. ACL 2020
l Naturalspeech 2: Latent diffusion models are natural and zero-shot speech and singing synthesizers. NeurIPS 2023
lMultilingual neural machine translation with knowledge distillation. ACL 2020
l NaturalSpeech: End-to-End Text-to-Speech Synthesis With Human-Level Quality. NeurIPS 2022
l Frank: a ranking method with fidelity loss. ACM SIGIR 2019
l Adaspeech: Adaptive text to speech for custom voice. Interspeech 2021
l Deliberation networks: Sequence generation beyond one-pass decoding. ACL 2021
l Understanding and improving transformer from a multi-particle dynamic system point of view. NeurIPS 2021
l Learning to teach. ICML 2017
l A study of reinforcement learning for neural machine translation. ACL 2016
l Supervised rank aggregation. ACM SIGKDD 2012
l Query dependent ranking using k-nearest neighbor. ACM SIGIR 2008
l Fully parameterized quantile function for distributional reinforcement learning. ICML 2020
主要成就与荣誉：
l 2017年以计算机科学家的身份荣获《北京青年》周刊 “年度匠人精神青年榜样” 奖项
l 提出的对偶学习助力微软在2018年中英新闻翻译任务上达到了人类专家水平
l 带领团队在WMT2019机器翻译大赛中获得8个项目的冠军
l 2019年设计了当时最高效的语音合成模型FastSpeech，实现了百倍的加速，并成为微软云Azure服务上支持100多种语言和200多种语音的基础模型组件。
l 2019开发了有史以来最强大的麻将AI Suphx，成为“天凤”平台上首个升至十段的AI，其稳定段位显著优于人类顶尖选手
l 2020年在国际知名的学术出版集团施普林格·自然（Springer Nature）出版了学术专著《对偶学习》
l 2022年发布了BioGPT模型，在生命科学领域大幅超越了其他大型语言模型，并在PubMed问答任务上首次达到了人类专家的水平
l 荣获ICDM 2022最佳学生论文亚军