AI for Research
Overview
Focus on the intersection of AI and science, build runnable algorithmic solutions or autonomous exploration environments around real scientific problems, and make AI an effective tool for scientific discovery.
This track focuses on the intersection of AI and science. It encourages participants to build runnable algorithmic solutions or autonomous exploration environments around real scientific problems, making AI an effective tool for scientific discovery.
The track is not only about single-point model performance. It focuses on how to turn scientific problems into computable, verifiable, and reproducible exploration tasks, forming a complete loop from problem definition and environment design to algorithm execution and interpretation of research signals.
Prize Information
*The winner of the Grand Award will be selected from the four track champion teams.
Additional Awards
Question Types
For scientific problems with clear evaluation methods, participants use datasets and evaluation frameworks provided by the organizing committee to model and solve the task, forming runnable and reproducible algorithmic solutions. Initial fields include virtual cells, small-molecule protein binding trajectory prediction, and materials science literature-driven scientific discovery agents.
For important scientific problems that have not yet been structured as standard evaluation tasks, participants are encouraged to define their own problems, exploration environments, and discovery signals, turning scientific intuition into an operating loop that agents can continuously explore.
Encourage teams to turn scientific problems into computable, verifiable, and reproducible exploration environments and algorithmic solutions. Negative results are allowed, but the process must be explainable, inspectable, and extensible.
Core Work Requirements
Core Work Requirements
- Entries may choose either algorithmic problems or open exploration problems, and should form runnable, reproducible, and verifiable solutions or exploration environments around real scientific problems.
- Entries should clearly disclose problem definition, evaluation or discovery signals, data sources, dependent tools, runtime flow, reproducibility method, and open-source plan. Open exploration entries also need to explain the minimum reference baseline and exploration log design.
- The preliminary round focuses on problem value, technical feasibility, and open-source potential. The semi-final requires a runnable environment or final code, technical documentation, and experiment / exploration results. The final focuses on live explanation, research signals, and long-term research potential.
Submission Requirements
Detailed submission requirements are as follows:
*The schedule may be adjusted flexibly according to actual progress and is subject to the committee’s final notice.
Review Focus
Projects will focus on the following areas:
- Problem definition and scientific value
- Exploration environment / evaluation framework design
- Runnability, reproducibility, and evidence quality
- Exploration process and research signals
- Open contribution and long-term research potential
Participant Support
This track will provide development resource support to participating teams that complete their preliminary-round submission and pass the submission validity review. The support may include computing resources, cloud services, or equivalent competition-related resources, with a maximum value of RMB 200 per team. Support will be allocated based on teams' overall preliminary-round scores, with no more than 300 allocations available per track in principle. After the preliminary-round submission deadline, the Organizing Committee will notify eligible teams via email or official competition community channels to submit the required materials, based on the submission and review results, and will distribute the resources accordingly. This resource support does not constitute a competition prize and will not affect evaluation scores or qualification for advancement. The specific forms of support, eligibility criteria, documentation requirements, distribution procedures, and timeline shall be subject to further notice from the Organizing Committee.
Best-fit Teams
- This track is suitable for interdisciplinary teams with AI / ML capabilities and real understanding of scientific fields such as biology, chemistry, materials, astronomy, and physics, as well as PhD students, postdoctoral researches, young PIs, university labs, and open-source teams hoping to explore AI solutions around unsolved scientific problems.