Scholar

Hao Zhou (周浩)

Google Scholar ID: XEgaCScAAAAJ

Insititute for AI Industry Research (AIR), Tsinghua University

LLMAI Scientist

Citations & Impact

All-time

Citations

6,581

H-index

i10-index

Publications

Co-authors

list available

Contact

Publications

20 items

Browse publications on Google Scholar (top-right) ↗

Resume (English only)

Academic Achievements

1. Seed Diffusion Preview: A large-scale language model based on discrete-state diffusion, achieving an inference speed of 2,146 token/s, outperforming Google's Gemini Diffusion and Mercury.
2. AMix-1: A systematic paradigm for crafting protein foundation models, emphasizing scalable protein design. Designed a high-activity AmeR variant with a 50× improvement over the wild type.
3. MemAgent: Demonstrated superb long-context capabilities, extrapolating from an 8K context trained on 32K text to a 3.5M QA task with <5% performance loss and 95%+ accuracy in 512K RULER test.
4. DAPO: An open-source large-scale reinforcement learning system, applied to Qwen2.5-32B base model, outperforming the previous state-of-the-art DeepSeek-R1-Zero-Qwen-32B on AIME 2024, achieving 50% accuracy with 50% fewer training steps.
5. Honors and Awards: Beijing Nova Program (2024), Ranked #1 on WMT EN-DE translation (2021), CCF NLPCC Distinguished Young Scientist (2021), CAAI Doctoral Dissertation Award (2019), Outstanding Doctoral Dissertation of Nanjing University (2019), National Scholarship (2011/2016).

Research Experience

Previously, he was a Research Scientist/Manager at Bytedance, where he led the ByteDance AI Lab research teams on text generation and AI drug design. Currently, he leads the GenSI research group at Tsinghua AIR and co-leads the SIA Lab, a joint laboratory of THU-AIR and Bytedance Seed, focusing on cutting-edge LLM research.

Background

Research Associate Professor at the Institute for AI Industry Research (AIR), Tsinghua University. His work focuses on Generative AI, particularly for discrete symbols (text and proteins). He specializes in Large Language Models (LLMs) and their applications to scientific discovery, with a special interest in biological agents.

Miscellany