A student-led international collaboration program connecting graduate students in data science across East Asia, mainly Japan and Korea.
Following the successful EAR-MP program in 2025, EAR-MP 2026 invites graduate students from member universities to collaborate on data-science projects. Faculty participation is intentionally minimized so that students can maximize autonomous teamwork and international exchange.
Complete the Google Form by August 21, 2026, 23:59 (JST/KST).
Open application formSelect a card to view the detailed project description.
Capture fine structures such as barbs, distinguish faint filament material from ground-based noise, and identify each filament as one coherent object. Classical image processing and deep learning approaches are both welcome.
MAGFiLO contains 10,244 manually annotated filaments across 1,593 GONG observations, with polygon masks, spines, bounding boxes, and chirality labels.
Evaluation includes IoU, precision, recall, AP@IoU, hit/miss rate, and the organizers’ Multi-scale IoU for fine structures.
Kaggle. Cash prizes and/or IEEE Big Data 2026 registration for top teams, with invited talks at the proposed SABiD workshop.
Azim Ahmadzadeh, Dustin Kempton, Qin Li, and Alexei Pevtsov.
Verifiable Chain Reasoning: solve symbolic, multi-step finance problems with automatically checked intermediate steps.
Market-Neutral Hedging: select asset pairs and manage dollar-neutral positions using prices, news, and filings.
Financial Audit Verification: verify reported values against XBRL calculation networks and the US-GAAP taxonomy of real SEC filings.
Built on FinChain and HERCULEAN, with hidden seeds, market windows, and newly collected filings to discourage memorization.
Answer accuracy and ChainEval; Sharpe Ratio, cumulative return, and maximum drawdown; audit accuracy and structural, extraction, and calculation errors.
Kaggle with an organizer-maintained Docker evaluation server for the agentic tasks.
The Fin AI with MBZUAI, McGill, Stevens, Yale, and the University of Manchester.
Generate synthetic LiDAR by any method—CARLA, diffusion or flow models, neural radiance fields, or hybrids—and train a 3D object detector without real labels.
LUMPI in Hannover, Germany, and V2X-Real in Los Angeles, USA.
Combined score of 0.6 detection and 0.4 realism, using KITTI 3D mAP, Chamfer Distance, MMD, Earth Mover’s Distance, and Fréchet Point-cloud Distance.
Codabench. USD 2,000 prize pool with per-track awards.
UrbanTwin reference pipeline, ready-made synthetic datasets on Harvard Dataverse, and the open-source LiGuard toolkit.
Muhammad Shahbaz and Shaurya Agarwal, Urbanity Lab, University of Central Florida.
Forecast seven annual carbon variables, including vegetation height, aboveground biomass, soil carbon, leaf area index, and productivity and respiration measures.
40 years of global data at 0.5° resolution, 136 input variables, 54,152 land sites, 15 forest-age conditions, and 812,280 forecasting sequences.
RMSE, MAE, cumulative error, and year-to-year delta error on a hidden test set of future ecosystem states.
Kaggle, with baselines and starter notebooks on GitHub.
Yiqun Xie, Zhihao Wang, Lei Ma, George Hurtt, Xiaowei Jia, and Yanhua Li.
Open Division for broad participation and Expert Division for full physical benchmarking.
TrafficStateBench, ODMEBench, and ShockwaveBench.
A reproducible California benchmark with 150 loop detectors across five Los Angeles freeways at 5-minute resolution for all of 2025, plus GPS traces, POI trip generation, and NGSIM trajectories.
Composite scores combining accuracy, physical consistency, demand attribution, congestion diagnostics, and reproducibility.
Codabench / EvalAI, cross-posted to Kaggle. Gold $1,500, Silver $1,000, Bronze $500, and Student/Newcomer $500 per division and track.
Registration opens July 15, 2026. Final submission November 6, 2026. Results workshop at IEEE Big Data 2026, Phoenix, December 14–17, 2026.
SUTPC, IEEE ITS Society TC, and RERITE, with collaborators including Cathy Wu and Yudai Honma.
Three long-horizon tracks inside χ-World: prior authorization, utilization management, and care management.
A high-fidelity simulator of 20 healthcare applications, 151 REST APIs, and 87 MCP tools.
About 5,000 chart activities for 50 simulated patients and about 90 healthcare workers, with no real patient data or PHI.
A deterministic contract checker and a rubric-based LLM judge evaluate terminal status, routing, payloads, required artifacts, and policy-grounded reasoning. Primary metric: pass@1.
Gold $1,500, Silver $1,000, Bronze $500, plus a Reliability Award. Total prize pool: $3,000.
Weiran Yao, Frank Wang, Haolin Chen, clinical advisors from Johns Hopkins Medicine and Wellstar Health System, and an international scientific advisory group.
Treat each detector as a plug-in black box producing time-stamped decisions with confidence values. Align decisions within sliding windows and combine them using agreement gating, confidence-weighted voting, or fast-proposes / slow-confirms cascades.
Build a reusable detector-agnostic fusion framework and test whether it reduces false alarms compared with individual detectors and naive synchronous fusion.
Public multivariate time-series streams with labeled anomalies from server, spacecraft, and infrastructure monitoring.
A per-timestamp anomaly verdict evaluated by point-wise and event-wise F1.
SMD, MSL, SMAP, and PSM benchmark datasets.
Yi Sangwan, Chonnam National University — salmonpoke20@jnu.ac.kr
Detailed information for Topics 1–6 is available through IEEE Big Data Cup 2026.
Each team will conduct the project through online meetings such as Zoom under the leadership of the project leader. An offline meeting may be held at the end of the project, subject to budget availability.
Project due date: December 15, 2026
Travel support to the IEEE Big Data Conference in Phoenix, Arizona, December 14–17, 2026, may be provided to winners and runners-up of IEEE Big Data Cup topics.
EAR-MP will select one winning team and two runner-up teams. Every member of the selected teams will be invited to an offline workshop in Korea or Japan in February 2027.
For questions, contact the steering committee member at your university or Prof. Ki-Joune Li at lik@pnu.edu.
| Name | Affiliation | |
|---|---|---|
| Saekwang Nam | Kyungpook National University | s.nam@knu.ac.kr |
| Yohei Shida | University of Tsukuba | shida@sk.tsukuba.ac.jp |
| Jae-woong Lee | Kangwon National University | jaewoong.lee@kangwon.ac.kr |
| Junhwa Chi | Pukyong National University | jchi@pknu.ac.kr |
| Inho Hong | Chonnam National University | ihong4867@gmail.com |
| Ki-Joune Li | Pusan National University | lik@pnu.edu |
| Tadashi Dohi | Hiroshima University | dohi@hiroshima-u.ac.jp |