Alibaba Group researcher in open-ended LLM agent RL — credit assignment and rubric-grounded self-evolution.
2
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—
The output axis is not scored: the snapshot provides no publication-count or venue-volume metrics for this researcher. It would measure total research production — number of papers, authorship roles, and venue spread over time.
Impact—
Impact is not scored: no citation counts, citation-derived indices, or influence metrics are available in the input data. This axis would measure how widely the researcher's work is cited and taken up by later studies.
Focus100/100
Focus scores 100, the maximum, and the underlying field distribution explains why: it contains exactly one field label, machine learning (LLM agents and reinforcement learning), and both catalog papers sit inside it. ArenaFlow tackles credit assignment for open-ended agent RL while ARISE-RL tackles rubric-based reward design and self-evolution — the same sub-area rather than scattered topics. The score rests on a single cataloged field entry, so depth is clear but breadth is untested.
Collaboration62/100
Collaboration scores 62, computed from 22 distinct coauthors in the snapshot. The frequent-coauthor list is anchored by names such as Pengjun Xie, Zheng-Jun Zha, Kaipeng Zhang, Boli Chen and Fanrui Zhang, consistent with large-team industrial research at Alibaba Group. On both reviewed papers Ding sits in a middle authorship position and is not the corresponding author, which fits a contributing role inside bigger teams rather than solo or lead work.
Activity—
Activity is not scored: the data lacks a year-by-year publication record, so recent output cadence cannot be assessed. This axis would measure publishing frequency and continuity across recent years.
Reproducibility55/100
Reproducibility averages 55, and the raw record shows this mean is computed over just one repro-checked paper (reproPapers: 1). Both catalog entries carry a medium repro-readiness tag: Evolve flags ArenaFlow for judge circularity and missing variance tests, and ARISE-RL for benchmarks that are self-built. No released code or model artifacts are recorded in the snapshot, so the score signals partial methodological transparency rather than verified replication.
Research summary
Ruixue Ding is a researcher affiliated with Alibaba Group (China), with this profile name-resolved from public bibliographic data under ORCID 0009-0006-8632-8032. The Evolve catalog holds two 2026 arXiv preprints, both in machine learning for LLM agents and reinforcement learning, with Ding listed as a middle author and not the corresponding author on either. ArenaFlow extends tournament-based trajectory ranking for open-ended agent RL with hierarchical, step- and skill-level credit propagation; Evolve's review records a +19.5 improvement over ArenaRL on Open-Travel while flagging judge circularity and missing variance tests. ARISE-RL proposes rubric-grounded generator-solver co-evolution with reward-gated self-distillation and claims state-of-the-art open-ended agent RL, though its benchmarks are self-built. The radar shows a highly concentrated focus: the field distribution contains a single label — machine learning (LLM agents and reinforcement learning) — giving a focus score of 100. Collaboration scores 62 on 22 distinct coauthors, with recurring partners including Pengjun Xie, Zheng-Jun Zha, Kaipeng Zhang and Boli Chen. Reproducibility averages 55 from one repro-checked paper, and both entries carry medium repro-readiness tags. Output, impact and activity are unscored for lack of data, making this an early-stage profile resting on two preprints.
Collaboration pattern
Collaboration is team-based industrial research: the snapshot records 22 distinct coauthors, with recurring names including Pengjun Xie, Zheng-Jun Zha, Kaipeng Zhang, Boli Chen, Fanrui Zhang, Xi Chen, Qiang Zhang, Shihang Wang, Yinfeng Huang, Yi Zheng and Jiawei Liu. Both reviewed papers place Ding in a middle authorship position without corresponding status, suggesting a contributing specialist role within larger Alibaba Group teams on LLM-agent RL projects.
Research keywords
Highlights
Hierarchical credit propagation for open-ended agent RL
Rubric-grounded self-evolution claims SOTA open-ended agent RL
Medium reproducibility readiness across both studies
Both papers carry a medium repro-readiness tag, and the reproducibility score (55) is averaged over one repro-checked paper. Evaluation-rigor flags — judge circularity, no variance tests, self-built benchmarks — are the main gaps.
Timeline
- 2026ARISE-RL posted to arXiv: rubric-grounded generator-solver co-evolution with reward-gated self-distillation, entering the Evolve catalog.
- 2026ArenaFlow posted to arXiv: hierarchical credit propagation over tournament-based ranking, reported at +19.5 over ArenaRL on Open-Travel.
- 2026Both catalog papers land in the same year within LLM agent RL, producing a perfectly concentrated field signature (focus score 100) with 22 distinct coauthors.
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 dimensions are computed from OpenAlex-based public bibliometric data only; output, impact and activity axes lack sufficient data and are not scored.
- The profile is name-resolved from public bibliographic records (ORCID 0009-0006-8632-8032); verify identity before citing statistics.
- Coverage is limited to papers reviewed in the Evolve catalog — two 2026 arXiv preprints — with no peer-review venue or citation information available.
- The frequent-coauthor list contains the name 'Xi Chen' twice, which may refer to two different individuals and is not disambiguated here.