HKUST researcher coupling LLM post-training (RL) and multimodal AI for physics; last author of two 2026 preprints.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-02) 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—
Unscored: no publication counts, venue record or authorship-volume data is available in this snapshot. This axis would measure total research output — number of papers, publication venues and yearly authorship roles — once a fuller bibliographic record exists.
Impact—
Unscored: citation counts, h-index and downstream adoption are absent from the provided data. This axis would measure how widely the work is cited, built upon and applied by other researchers across venues.
Focus0/100
Scores 0/100 because the field distribution is an even 50/50 split: half machine learning for LLM post-training and reinforcement learning, half AI-for-physics and scientific multimodal learning. The two 2026 cataloged papers sit on opposite sides of this divide, so the portfolio spans two distinct agendas rather than one concentrated research line.
Collaboration53/100
Scores 53/100. The snapshot records 13 distinct coauthors, with Henan Sun, Haitao Hu, Jianfeng Zhang and Nuo Chen recurring across the cataloged work. Jia Li holds the last-author slot on both papers while being non-corresponding author, a pattern consistent with a senior supervisory role within mid-sized, stable teams.
Activity—
Unscored: no yearly publication cadence or recent-output record is provided beyond the two 2026 preprints. This axis would measure publishing tempo and continuity over recent years once a complete yearly record is available.
Reproducibility20/100
Scores 20/100, the mean across two assessed papers (reproPapers = 2). Both carry low reproducibility readiness: the DCRL review flags thin theory and missing key ablations, while PhyMo reports promising gains with no released code, so independent replication of either result would currently demand substantial extra effort.
Research summary
Jia Li is a Hong Kong University of Science and Technology–affiliated machine-learning researcher whose cataloged work runs on two parallel tracks: reinforcement-learning-based post-training for large language models, and multimodal AI for physics. Both entries in the Evolve catalog are 2026 arXiv preprints with Jia Li in the last-author, non-corresponding position — a placement that typically signals a senior or supervising role. The first, DCRL, proposes decoupling-and-coupling reinforcement learning via policy-reward manifold alignment, co-evolving policy and reward through shared weights and evolving rubrics; Evolve's review credits a 4B model outperforming a 32B one while flagging thin theory and missing ablations. The second, PhyMo, casts PDE-structured physical fields as a new modality for multimodal AI4Physics, showing multi-dataset gains that reviewers judge promising but with thin margins, trivial theory and no code. Bibliometric raw data records 13 distinct coauthors, with Henan Sun, Haitao Hu, Jianfeng Zhang and Nuo Chen recurring, and an even 50/50 field split across the two tracks — yielding focus 0, collaboration 53 and reproducibility 20 on the radar. Output, impact and activity axes lack underlying data and remain unscored. As the name is shared by many researchers, this profile is name-resolved from public bibliographic data.
Collaboration pattern
The snapshot records 13 distinct coauthors, with Henan Sun, Haitao Hu, Jianfeng Zhang and Nuo Chen recurring across the cataloged work, joined by Zehua Li, Qifan Zhang, Jin Liu and Lujia Pan. Holding the last-author, non-corresponding slot on both 2026 preprints suggests a senior, supervisory position within compact, repeat teams; the collaboration dimension scores 53/100, a mid-range pattern.
Research keywords
Highlights
DCRL: co-evolving policy and reward via policy-reward manifold alignment
PhyMo: PDE-structured physical fields as a new modality for AI4Physics
Senior (last) author across two distinct 2026 research lines
Both cataloged 2026 preprints place Jia Li in the last-author, non-corresponding position — one on LLM post-training with reinforcement learning, the other on scientific multimodal learning for physics — evidencing a supervisory role spanning two agendas.
Timeline
- 2026DCRL preprint posted to arXiv, proposing decoupling-and-coupling reinforcement learning with policy-reward manifold alignment; Jia Li in the last-author position.
- 2026PhyMo preprint posted to arXiv, introducing PDE-structured physical fields as a modality for multimodal AI4Physics; again in the last-author position.
- 2026Both preprints enter the Evolve review catalog, each assessed with low reproducibility readiness.
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.
- "Jia Li" is a common name and no ORCID is provided; this profile is name-resolved from public bibliographic data and may not uniquely identify the individual.
- Radar scores derive from a small bibliometric snapshot; the output, impact and activity axes have insufficient underlying data and are unscored.
- The Evolve catalog contains only two papers for this researcher, both 2026 arXiv preprints; no peer-reviewed venue or citation data is included.
- The reproducibility score (mean 20) rests on just two assessed papers and may shift as more of this researcher's work is reviewed.