Alibaba researcher in LLM post-training: RL-based distillation and curriculum methods for instruction following
2
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
0
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 scoreable: the snapshot provides no reliable cumulative work count for this identity, and h-index and i10-index are recorded as 0. This axis would measure total publications, authorship-position share, and yearly output volume.
Impact—
Not yet scoreable: citation totals are absent, and h-index (0) and i10-index (0) give no usable signal at this thin OpenAlex coverage. This axis would measure citation accumulation and uptake of the work by later research.
Focus100/100
Focus scores 100, driven by fieldDist: the only field-labeled work in the snapshot is 'machine learning (LLM post-training)', so concentration is maximal by construction from a single labeled paper rather than a long track record. Topic tags hint at wider activity — Topic Modeling (7), Reinforcement Learning in Robotics (4), Domain Adaptation and Few-Shot Learning (4), Natural Language Processing Techniques (3), Information Retrieval and Search Behavior (2) — that the sparse field labels do not yet resolve.
Collaboration55/100
Collaboration scores 55, based on 15 distinct coauthors in the snapshot. Work concentrates around a stable industrial cluster: the frequent-coauthor list repeats ten names — Yanzhao Zheng, Hangcheng Zhu, Ruohui Huang, Zhentao Zhang, Tianze Xu, Chao Ma, Jihuai Zhu, Jiashun Liu, Xinle Deng, and Baohua Dong — with Tianze Xu and Chao Ma each appearing twice, suggesting a fixed in-house team rather than loose academic networks.
Activity67/100
Activity scores 67 from recent3yWorks of 9 and countsByYear showing 1 work in 2025 rising to 8 in 2026. Both cataloged papers are 2026 arXiv releases — one middle-author, one first-author (PAC) — so the current output surge is recent and concentrated in LLM post-training; longer-term cadence cannot yet be judged.
Reproducibility55/100
Reproducibility averages 55 from exactly 1 Evolve-reviewed paper (reproPapers: 1, reproMean: 55). That paper, CC-OPD, is rated medium readiness: the review credits best average scores on 7 benchmarks across 2 Qwen model pairs, but flags single-seed runs and no confidence intervals. A one-paper base makes this score provisional; PAC awaits review and could move it.
Research summary
Yuanqiang Yu is a researcher at Alibaba Group (China) whose public bibliographic footprint, resolved via OpenAlex (identity: openalex; ORCID 0000-0002-8697-883X), is early-stage and rapidly growing. The snapshot records 9 works in the last three years, concentrated in 2026 (8 works, versus 1 in 2025). Two 2026 arXiv papers are in the Evolve catalog: a middle-author contribution to 'Counterfactual Constraint-Conditioned On-Policy Distillation for Multi-Constraint Instruction Following' (CC-OPD), reviewed with medium reproducibility readiness — best average on 7 benchmarks over 2 Qwen pairs, though single-seed runs and no confidence intervals warrant caution — and a first-author paper, 'PAC: Progress-Augmented Advantage Curriculum for Multi-Task Reinforcement Learning of LLMs', which awaits review. Field labels are sparse (one labeled field: machine learning, LLM post-training), yielding a maximal focus score of 100, while topic tags — led by Topic Modeling (7), Reinforcement Learning in Robotics (4), and Domain Adaptation and Few-Shot Learning (4) — indicate broader activity. Collaboration is moderate (score 55) around 15 distinct coauthors, with a recurring industrial team. Citation-based axes (output, impact) are unscoreable: h-index and i10-index read as 0, reflecting thin coverage rather than measured non-impact. Given the common name, this profile should be read as name-resolved from public bibliographic data.
Collaboration pattern
Collaboration is anchored in an industrial research team: the snapshot lists 15 distinct coauthors (score 55), and the frequent-coauthor list repeats ten names — Yanzhao Zheng, Hangcheng Zhu, Ruohui Huang, Zhentao Zhang, Tianze Xu, Chao Ma, Jihuai Zhu, Jiashun Liu, Xinle Deng, and Baohua Dong — with Tianze Xu and Chao Ma appearing twice. No external-academia partners are visible in the provided data, so the pattern reads as a stable in-house LLM post-training group.
Research keywords
Highlights
CC-OPD: counterfactual reward shaping for multi-constraint instruction following
First-author PAC curriculum for multi-task RL of LLMs
First-authored 2026 arXiv paper, 'PAC: Progress-Augmented Advantage Curriculum for Multi-Task Reinforcement Learning of LLMs'. It is cataloged but not yet reviewed — no Evolve verdict, keywords, or repro-readiness are available, so its claims await assessment.
Output surge in 2026 around LLM post-training
countsByYear shows 1 indexed work in 2025 rising to 8 in 2026 (recent3yWorks: 9), and the only fielded label is machine learning (LLM post-training) — a steep, recent ramp concentrated in one direction.
Timeline
- 2025First year with indexed output in this snapshot: OpenAlex records 1 work in 2025.
- 2026Indexed output rises to 8 works — the most active year recorded in this profile.
- 2026Middle-author co-author on 'Counterfactual Constraint-Conditioned On-Policy Distillation for Multi-Constraint Instruction Following' (arXiv), the paper Evolve reviewed at medium reproducibility readiness.
- 2026First author on 'PAC: Progress-Augmented Advantage Curriculum for Multi-Task Reinforcement Learning of LLMs' (arXiv), now in the Evolve catalog pending review.
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 are computed from OpenAlex records only (identity: openalex); no other citation index is reflected.
- Output and impact axes are missing: h-index and i10-index read as 0, likely reflecting thin OpenAlex coverage rather than measured values.
- Only one cataloged paper (CC-OPD) has a completed Evolve review; PAC has no verdict, keywords, or repro-readiness yet.
- 'Yuanqiang Yu' is a common Chinese name; this profile is name-resolved from public bibliographic data (ORCID 0000-0002-8697-883X) and may mix records from different people.