NLPCC 2026 SHARED TASK
AISB
AI Scientist Benchmark
Evaluating autonomous AI research capability — discover problems, think, explore, validate through experiments, and communicate findings.
评测AI自主科研能力 — 发现问题、思考探索、假设-实验验证、将结论讲出来让人理解
3
Research Directions
研究方向
2
Competition Tracks
竞赛赛道
24
Reference Papers
参考论文
CAS
Integrity Audit
诚信审核
The Research Cycle
发现问题 → 思考探索 → 实验验证 → 表达结论
Track A 以科学发现为核心，Track B 以实验提升为核心，共享 AI Scientist 底座
Competition Tracks
两大赛道，全面评估 AI 科研能力
TRACK 1
Scientific Research
科学研究赛道
Discover scientific problems from literature, form hypotheses, validate through experiments, and write a paper that communicates the findings clearly. Experiments support the idea — the idea is the core.
TRACK 2Benchmark SOTA Challenge
基准提升赛道
Propose scientifically grounded methods to improve SOTA. Experiments are the primary evidence, but you must explain why your method works — interpretable improvement over blind optimization.
Key Dates / 重要日期
Jun
Announcement
6月 · 竞赛公告
Jul
Task Packages
7月 · 任务发布
Oct 1
Deadline
10月1日 · 提交截止
Nov
NLPCC Macau
11月 · 澳门发布