8 papers
Science Earth: Towards A Planet-Scale Operating System for AI-Native Scientific Discovery
Zhe Zhao, Haibin Wen, Yingcheng Wu +10
Scientific discovery demands intelligence, perseverance, and serendipity across vast search spaces. Today, top scientific capabilities remain siloed--one AI system for biological a…
ResearchEVO: An End-to-End Framework for Automated Scientific Discovery and Documentation
Zhe Zhao, Haibin Wen, Jiaming Ma +4
An important recurring pattern in scientific breakthroughs is a two-stage process: an initial phase of undirected experimentation that yields an unexpected finding, followed by a r…
Autogenesis: A Self-Evolving Agent Protocol
Wentao Zhang, Zhe Zhao, Haibin Wen +4
LLM-based agent systems can adapt by revising prompts, constructing tools, and changing other runtime resources. Managing these heterogeneous updates requires a shared lifecycle fo…
Logos: An evolvable reasoning engine for rational molecular design
Haibin Wen, Zhe Zhao, Fanfu Wang +4
The discovery and design of functional molecules remain central challenges across chemistry,biology, and materials science. While recent advances in machine learning have accelerat…
EvoCodeBench: A Human-Performance Benchmark for Self-Evolving LLM-Driven Coding Systems
Wentao Zhang, Jianfeng Wang, Liheng Liang +3
As large language models (LLMs) continue to advance in programming tasks, LLM-driven coding systems have evolved from one-shot code generation into complex systems capable of itera…
Evolution without an Oracle: Driving Effective Evolution with LLM Judges
Zhe Zhao, Yuheng Yang, Haibin Wen +3
The integration of Large Language Models (LLMs) with Evolutionary Computation (EC) has unlocked new frontiers in scientific discovery but remains shackled by a fundamental constrai…
From Understanding to Excelling: Template-Free Algorithm Design through Structural-Functional Co-Evolution
Zhe Zhao, Haibin Wen, Pengkun Wang +8
Large language models (LLMs) have greatly accelerated the automation of algorithm generation and optimization. However, current methods such as EoH and FunSearch mainly rely on pre…
Delayed Bottlenecking: Alleviating Forgetting in Pre-trained Graph Neural Networks
Zhe Zhao, Pengkun Wang, Xu Wang +5
Pre-training GNNs to extract transferable knowledge and apply it to downstream tasks has become the de facto standard of graph representation learning. Recent works focused on desi…