Tiangang Li

Wuhan University · China

Unclaimed · AI-structured profileORCID OpenAlex

Wuhan University researcher applying deep RL and graph representations to cloud and microservice scheduling

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Insufficient data to score. This axis measures total scholarly output over time — work counts, venue mix, and productivity trajectory. The snapshot confirms only 12 recent three-year works and sparse yearly counts, too thin for a reliable rating.

Impact—

Insufficient data to score. This axis measures citation-based influence — total citations, h-index trajectory, and uptake by other researchers. The snapshot holds only h-index 3 and i10-index 2 on a common-name record likely affected by entity mixing, so no impact rating is given.

Focus0/100

Scored 0/100. The indexed field distribution splits evenly (0.5/0.5) between cloud computing systems / deep reinforcement learning scheduling and cloud computing (ML-based systems scheduling) — two closely aligned areas — yet OpenAlex topic clusters also surface congenital heart disease studies, security and verification in computing, adversarial robustness, and digital media forensic detection (3 works each), a scattered topical footprint that lowers measured concentration and likely reflects name-disambiguation noise on this common-name record.

Collaboration32/100

Scored 32/100. Only 4 distinct coauthors appear in the record — Shi Ying, Xiangbo Tian, Chuan Shi, and Ding Xiao as recurring collaborators — a compact network relative to 12 works in the last three years. Both reviewed 2026 papers are first-author, non-corresponding works, consistent with a lead-executor role inside one stable team rather than a broad, cross-institution collaboration web.

Activity75/100

Scored 75/100. Activity is this record's strongest signal: 12 works in the last three years and a rising annual curve — 2 works in 2024, 4 in 2025, and 6 in 2026, the highest year recorded. Earlier years (2005, 2010, 2014, 2021) are sparse and intermittent, so the current cloud-scheduling period marks a clear step-up in sustained productivity.

Reproducibility55/100

Scored 55/100, the mean over 2 Evolve-reviewed papers, both rated medium reproducibility readiness. HiGFRL shows solid design and ablations but no variance testing and simulation-only evaluation; MCRL2 reports single-run results in trace-driven simulation, with its cross-node claim flagged as shaky. Adding variance reporting, multiple seeds, and real-cluster validation would be the clearest path to lifting this axis.

Research summary

Tiangang Li is a researcher at Wuhan University whose recent work centers on learning-based scheduling in cloud computing systems. Evolve has reviewed two 2026 first-author arXiv preprints: HiGFRL, a hierarchical graph fusion-driven reinforcement learning scheduler for dependency-aware DAG task scheduling in heterogeneous clouds, and MCRL2, a multi-resource cross-attention-augmented soft actor-critic method for cloud microservice scheduling. Both place the author at the intersection of deep reinforcement learning, graph and attention-based representation learning, and systems-level resource management. The OpenAlex footprint reads as active but early-stage: h-index 3, i10-index 2, 12 works in the last three years, and yearly output climbing from 2 works in 2024 to 4 in 2025 and 6 in 2026. The indexed field distribution splits evenly between cloud computing systems / deep reinforcement learning scheduling and ML-based cloud systems scheduling, although topic clusters also surface congenital heart disease studies, security and verification, adversarial robustness, and digital media forensics — a breadth that likely reflects name-disambiguation noise on a common-name record. Both reviewed papers carry medium reproducibility readiness (mean 55/100), with solid designs and ablations but simulation-only, single-run evaluation. Citation-impact metrics are not yet scored in this profile.

Collaboration pattern

The record shows a compact collaboration footprint: only 4 distinct coauthors, with Shi Ying, Xiangbo Tian, Chuan Shi, and Ding Xiao as recurring partners — a stable single-team pattern. Both reviewed 2026 papers are first-author, non-corresponding works, suggesting Li drives technical execution and writing while a senior colleague likely anchors supervision. The collaboration score of 32/100 reflects limited network breadth rather than absence of teamwork; cross-institution co-authorship would raise it.

Research keywords

cloud task schedulingDAG task schedulingdeep reinforcement learninggraph neural networkshypergraph convolutionheterogeneous cloud computingmicroservice schedulingload balancingsoft actor-critic (SAC)actor-critic (PPO)cross-attentionrepresentation learning

Highlights

HiGFRL: hierarchical graph fusion RL for dependency-aware scheduling in heterogeneous clouds
First-author 2026 arXiv preprint combining hypergraph and multi-view graph neural networks with PPO-based deep reinforcement learning to schedule DAG tasks on heterogeneous cloud resources. Evolve's review credits a solid design and thorough ablations, but notes the absence of variance testing and a simulation-only evaluation; reproducibility readiness is medium.
MCRL2: cross-attention-augmented SAC for cloud microservice scheduling
First-author 2026 arXiv preprint augmenting soft actor-critic with multi-resource cross-attention representation learning for microservice scheduling and load balancing. Evolve finds it beats DRL baselines in trace-driven simulation, but flags single-run results and a shaky cross-node claim; reproducibility readiness is medium.

Timeline

2005Earliest indexed work in the OpenAlex record (1 work); early coverage is sparse, so this anchor is weak.
2021A five-work year, the largest output spike before the current run.
20242 indexed works this year, beginning the current three-year ramp-up.
20254 works, doubling the prior year and continuing the ramp-up.
20266 works, the record's highest year; includes two first-author arXiv preprints reviewed by Evolve — HiGFRL and MCRL2.

Papers in the Evolve catalog

2026-09-14first author
HiGFRL: Hierarchical Graph Fusion-Driven Reinforcement Learning for Dependency-Aware Task Scheduling in Heterogeneous Cloud
2026-09-11first author
MCRL2: Multi-resource Cross-attention-based Representation Learning-augmented Reinforcement Learning for Cloud Microservice Scheduling

Frequent co-authors

Shi Ying×2 Xiangbo Tian×2 Chuan Shi×1 Ding Xiao×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometric indicators (h-index 3, i10-index 2, works per year, topics, coauthor counts) come from OpenAlex only and may be incomplete for this author.
The name is common; indexed topics outside cloud computing (congenital heart disease, security and verification, digital media forensics) may belong to homonymous researchers. The profile is name-resolved from public bibliographic data, not verified biography.
Output and impact dimensions are unscored because verified citation and venue data are insufficient in this snapshot.
Only two papers — both 2026 arXiv preprints with simulation-only evaluation — have been reviewed by Evolve; verdicts may change as peer-reviewed versions appear.