Yi Liu — graph learning and LLM-agent researcher, Chongqing University of Posts and Telecommunications
3
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Not yet scored: the current snapshot has no publication counts or venue records, so total output cannot be verified; this axis would measure annual paper volume and the venues where the work appears.
Impact—
Not yet scored: no citation counts or uptake metrics exist for the reviewed arXiv papers; this axis would track citations and downstream adoption of the hypergraph and agent methods.
Focus100/100
Focus reaches the maximum 100/100: the raw field distribution records a single field, machine learning (graph representation learning). The three reviewed papers, though split between hypergraph learning and LLM-agent methods, all target structured representations and reasoning behavior, suggesting a coherent methodological core rather than scattered topics.
Collaboration62/100
The radar scores collaboration 62/100. The raw record lists 21 distinct coauthors, with 12 recurring partners — including Wei Hu, Rongzhi Zhu, and Xiangyu Liu — appearing across both the hypergraph-learning and LLM-agent paper lines. All three reviewed papers place Yi Liu in a middle author position rather than as corresponding author, consistent with contributing inside moderate-sized teams rather than leading or solo work.
Activity—
Not yet scored: only three 2026 entries are on file, so recent publishing cadence cannot be established; this axis would measure annual publication rhythm and research continuity.
Reproducibility20/100
Reproducibility is the weakest recorded axis at 20/100 (reproMean 20, from 1 assessed paper). The granular-ball hypergraph paper is rated low repro-readiness due to weak statistics, specification gaps, and template leftovers, while the two agent papers rate medium; code and data availability remain unconfirmed in the reviewed material.
Research summary
Yi Liu is a researcher at Chongqing University of Posts and Telecommunications whose reviewed work connects graph-structured machine learning with LLM-based agents. The Evolve catalog holds three 2026 arXiv papers, in each of which Yi Liu appears in a middle author position rather than as corresponding author. The first applies granular-ball computing to hypergraphs, proposing multi-granularity adaptive hypergraph representation learning with adaptive hyperedge generation for node classification and claiming state-of-the-art results on 7 benchmarks; Evolve flags weak statistics, specification gaps, and template leftovers, rating repro-readiness low. The second introduces FactAgent, which converts sparse outcome rewards into Bayesian fact-utility step-level rewards for LLM search agents and reports 51.2 average exact match with a 7B model; Evolve calls the dense supervision promising but notes per-dataset claims outrun the evidence. The third identifies a linearly decodable 'late-stage pressure' state linked to premature submissions in long-horizon tool-use agents, with solid controls but modest gains and LLM-defined labels. The bibliometric snapshot records 21 distinct coauthors and 12 frequent collaborators, yielding a collaboration score of 62/100, a maximal focus score of 100/100, and a low reproducibility score of 20/100; output, impact, and activity are unscored for missing data. Because 'Yi Liu' is a common name, this profile is name-resolved from public bibliographic data.
Collaboration pattern
Across all three reviewed papers Yi Liu appears in a middle author position and never as corresponding author, pointing to contributions within larger teams. The record lists 21 distinct coauthors and a stable core of 12 frequent collaborators — including Wei Hu, Rongzhi Zhu, and Xiangyu Liu — spanning both the hypergraph-learning and LLM-agent lines. The radar collaboration score of 62/100 reflects a moderately broad but team-anchored network.
Research keywords
Highlights
Granular-ball hypergraph learning with claimed SOTA on 7 benchmarks
FactAgent: Bayesian fact-utility rewards for search agents
A linearly decodable 'late-stage pressure' state in tool-use agents
Timeline
- 2026Co-authored 'Multi-granularity Adaptive Hypergraph Representation Learning via Granular-ball' (arXiv), applying granular-ball computing to hypergraph learning in a middle author position.
- 2026Co-authored 'Dense Process Supervision for Search Agents via Fact Utility Estimation' (arXiv), introducing FactAgent with Bayesian fact-utility step rewards for LLM search agents.
- 2026Co-authored 'Polished but Unresolved: Identifying Late-Stage Pressure States in Long-Horizon Tool-Use Agents' (arXiv), an interpretability study of premature agent submissions.
Papers in the Evolve catalog
Frequent co-authors
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 scores derive from a limited OpenAlex-based snapshot; output, impact, and activity are unscored because the underlying data is missing.
- 'Yi Liu' is a common name; this profile is name-resolved from public bibliographic data and may mix homonymous authors.
- Coverage is limited to three 2026 arXiv papers reviewed by Evolve; no ORCID or country field was available to help confirm identity.
- The reproducibility score (reproMean 20) is based on only 1 assessed paper, so the 20/100 figure carries wide uncertainty.