Yexiang Xue

Research profile

Research summary

Yexiang Xue is a researcher at Purdue University, West Lafayette; the identity here is resolved from public bibliographic data (OpenAlex; ORCID 0000-0002-4533-0543). The record shows an h-index of 19, an i10-index of 35, and 33 works in the most recent three years, with per-year output tracked continuously from 2012 (2 works) through 2026 (16 works), peaking at 17 in 2023. Activity is the strongest axis (100/100), while the field distribution splits evenly: 50% machine learning — robot learning / imitation learning and 50% computing education / AI education outreach. Topics span reinforcement learning in robotics (14), machine learning in materials science (11), data classification (11), topic modeling (10), and Bayesian modeling and causal inference (10). Two 2026 arXiv papers in the Evolve catalog, both with Xue as last author, capture this dual track: 'Missing Bridges' introduces composition-aware active imitation learning that requests bridge demonstrations maximizing start-goal connectivity — reported 100% success with 3 demonstrations, though external validity is untested — and 'Teaching AI, Robotics, & Community' proposes a hubs-based K-12 robotics mentorship framework for rural schools. Frequent coauthors include Maxwell J. Jacobson, Ahmed H. Qureshi, Gustavo Rodriguez-Rivera, and Petros Drineas, with 6 distinct coauthors in the recent window and a collaboration score of 39. Reproducibility readiness averages 55 across the two reviewed papers.

Research keywords

Highlights

• Composition-aware active imitation learning with bridge demonstrations

  In the 2026 arXiv paper 'Missing Bridges' (last author), the AALT approach requests bridge demonstrations that maximize start-goal connectivity in latent topology, enabling compositional behavior with diffusion policies. Evolve's review calls the theory clean and notes striking reported gains — 100% success with 3 demonstrations — while flagging untested external validity. Reproducibility readiness: medium.

• Hubs-based K-12 robotics framework reaching rural schools

  The 2026 arXiv paper 'Teaching AI, Robotics, & Community' (last author) proposes ARC, a hubs-based K-12 robotics mentorship framework built on near-peer mentorship and service learning for rural STEM access. Evolve notes a promising but tiny single-site trial and an elaborate 40-year spatial Markov simulation whose optimistic projections rest on assumed rather than estimated parameters. Reproducibility readiness: medium.

Timeline

1. 2012Earliest publications in the provided counts-by-year record appear, with 2 works logged — the start of a continuous publication run tracked through 2026.
2. 2016Annual output climbs to 9 works and holds at 9-10 through 2018, marking the ramp-up phase of the record.
3. 2019Output reaches 15 works, then stays within the 7-17 range every year through 2026 — no gap years appear in the provided data.
4. 2023The highest single-year count in the record: 17 works, flanked by 14 in 2022 and 10 in 2024.
5. 2025A low of 7 works is logged, the smallest annual count since 2015, before output rebounds sharply.
6. 202616 works are logged, and two arXiv preprints enter the Evolve catalog with Xue as last author: 'Missing Bridges' (active imitation learning) and 'Teaching AI, Robotics, & Community' (K-12 education framework).

Papers in the Evolve catalog

• 2026-09-17

  Teaching AI, Robotics, & Community: A Hubs-Based K-12 Education Framework for Reaching Rural Schools

• 2026-09-17

  Missing Bridges: Composition-Aware Active Imitation Learning

Frequent co-authors