Early-career researcher on LLM agent evaluation, behavioral game theory, and personalized learning path benchmarks
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-20) 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 citation-based output measures are empty in this snapshot (h-index 0, i10-index 0, only 2 works recorded in 2026). This axis would track publication volume, venue quality and yearly productivity once a fuller publication record accumulates.
Impact—
Not yet scored: citation and influence data are absent here (h-index and i10-index both 0, no citation counts by year). This axis would measure how widely the two 2026 preprints are cited and taken up by follow-up work, and cannot be estimated from current data.
Focus0/100
The focus score is 0, the strongest dispersion signal this scale allows. The record spreads across five OpenAlex topics with one work each — reinforcement learning in robotics, game theory and applications, online learning and analytics, intelligent tutoring systems, and language and cultural evolution — while the field distribution splits evenly: 0.5 to multi-agent LLM evaluation and behavioral game theory, 0.5 to NLP / AI in education. Both strands share an LLM-evaluation core.
Collaboration51/100
Collaboration scores 51 on the 0-100 scale, grounded in 12 distinct coauthors recorded in the snapshot. Both cataloged papers are first-authored (non-corresponding) arXiv preprints, pointing to an active but very early coauthor network; the frequent-coauthor list spans names such as Wenwen Li, Yifan Dou, Kehai Chen and Min Zhang, though it also contains a same-name entry that signals entity-resolution noise.
Activity32/100
Activity scores 32, reflecting a very short publication tail: the snapshot counts only 2 works in the recent three-year window, both in 2026 (countsByYear lists 2 for 2026 and nothing earlier). This pattern is consistent with an early-stage profile whose record is just beginning rather than a slowdown from a longer history.
Reproducibility55/100
Reproducibility scores 55, computed from Evolve reviews of 2 papers (reproMean 55, reproPapers 2). Both preprints are rated medium repro-readiness: the multi-agent ICL study rests on LLM-agent simulations whose gains the authors themselves stress-test, while PersonaPath ships a large benchmark (2,000 personas, 347 textbooks) but is evaluated only in simulation.
Research summary
Yu Liu is an early-career researcher working at the intersection of large language model (LLM) evaluation and behavioral game theory, with a second strand in AI for education. The Evolve catalog lists two 2026 first-author arXiv preprints; the bibliometric snapshot records 2 works in the recent three-year window (both in 2026), 12 distinct coauthors, and an h-index and i10-index of 0 — a publication record at its very beginning. The first paper asks whether in-context learning in multi-agent interdependent decision-making reflects recursive reasoning or statistical extrapolation: using a rational-expectations-equilibrium diagnostic in public-goods and global-game settings, the Evolve review finds that ICL gains vanish when price trends break, favoring the statistical account, while cautioning that the reasoning null is not airtight. The second, PersonaPath, introduces a knowledge-centric benchmark for personalized learning path planning built from 2,000 personas and 347 textbooks with prerequisite relations, where the best model passes only 29.5% under simulation-only evaluation. The record splits evenly (0.5/0.5) across two adjacent fields, both sharing an LLM-evaluation core, and both papers carry a medium reproducibility readiness (mean 55).
Collaboration pattern
The snapshot records 12 distinct coauthors, supporting a collaboration score of 51. Both cataloged papers are first-authored but non-corresponding arXiv preprints, so the network is broad yet junior-stage. Recurring names include Wenwen Li, Yifan Dou, Guangnan Ye, Zeming Liu, Tianle Zhang, Kehai Chen and Min Zhang; the list also contains a same-name entry (Yu Liu), a known artifact of name-based entity resolution that warrants care when reading coauthor ties.
Research keywords
Highlights
In-context learning gains vanish when price trends break
PersonaPath: a knowledge-centric benchmark for personalized learning paths
Timeline
- 2026Published a first-author arXiv preprint asking whether in-context learning in multi-agent interdependent decision-making reflects recursive reasoning or statistical extrapolation — one of only 2 works recorded for 2026.
- 2026Released PersonaPath, a knowledge-centric benchmark for personalized learning path planning built from 2,000 personas and 347 textbooks, with the best model passing 29.5%.
- 2026Both preprints entered the Evolve catalog with medium repro-readiness (reproPapers 2, reproMean 55), alongside a snapshot showing 12 distinct coauthors and h-index and i10-index of 0.
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.
- Bibliometrics are derived from OpenAlex only; no institutional, publisher or citation-aggregator data beyond that source is included.
- "Yu Liu" is a very common name: this profile is name-resolved from public bibliographic data, and the frequent-coauthor list contains a same-name entry, indicating possible entity-resolution noise.
- No institution, country or ORCID is available in the snapshot, so affiliation-level context cannot be verified.
- Scope covers 2 arXiv preprints from 2026; the output and impact axes are unscored (insufficient citation data), with h-index and i10-index both 0.