Chinese Academy of Sciences researcher working on multimodal AI agents, RLVR post-training, and educational AI.
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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—
This axis measures total publication volume and productivity over time. It is not yet scored: the current snapshot lists only the three Evolve-reviewed 2026 arXiv preprints and provides no full publication count or venue history, so any output estimate would be speculative.
Impact—
Impact — typically citation counts, h-index, and citing literature — cannot be computed from the current data, which includes no citation statistics for the researcher or for the three reviewed 2026 preprints; this axis is therefore left unscored rather than estimated.
Focus0/100
The focus score is 0, and the raw field distribution explains why: output splits almost perfectly evenly — about 33% each — across multimodal AI agents, NLP/educational AI, and LLM post-training (RLVR). With all three cataloged papers dated 2026, this reads as an exploratory, evenly spread portfolio rather than an established specialization; more papers would be needed to know whether one thread becomes dominant.
Collaboration54/100
With 14 distinct coauthors and a collaboration score of 54, Chen works within a sizeable but not vast network. All three cataloged papers are middle-authored and non-corresponding, indicating contribution inside team-led projects rather than independent leadership; recurring names such as Wentao Zhang, Ran He, Hao Liang, and Hengyi Feng suggest a stable core group behind the 2026 output.
Activity—
This axis would track publishing cadence — papers per year and recency of output. It is unscored here: the snapshot contains only three 2026 preprints and no multi-year publication history, so temporal activity patterns cannot yet be characterized honestly.
Reproducibility70/100
Reproducibility scores 70, the mean readiness across exactly 3 reviewed papers. Consistent with that figure, DataFlex-RL is rated high — a 591-run matched-seed study built around explicit seed control — while VLM-in-Sandbox and OmniEdu are rated medium, with noted gaps such as statistically inconclusive gains and missing ablations. The pattern points to deliberate evaluation and seed-control discipline.
Research summary
Mingrui Chen is a researcher affiliated with the Chinese Academy of Sciences whose reviewed work spans three roughly equal strands: multimodal AI agents, LLM post-training with verifiable rewards, and educational AI — each accounting for about one third of the cataloged papers. All three entries in Evolve's catalog are 2026 arXiv preprints on which Chen is a middle-listed, non-corresponding author, consistent with a contributor role within a larger collaborative group. The DataFlex-RL paper reports a 591-run matched-seed null study concluding that RLVR data policies add no reproducible gain over uniform GRPO, and it carries Evolve's highest reproducibility readiness rating (high). VLM-in-Sandbox introduces a training-free visual workspace for sandboxed vision-language agents, with evaluation described as rigorous and honest even though the promote mechanism's accuracy gain is statistically inconclusive. OmniEdu applies capability-balanced SFT to Qwen 4-27B models across K-12 benchmarks, with gains judged plausible but tempered by overlapping train/eval sources and missing ablations. The methodological signature across the set is evaluative: platforms, controlled comparisons, and explicit reproducibility discussion. Bibliometrically, the reproducibility dimension scores 70 (mean over 3 papers), collaboration scores 54 with 14 distinct coauthors, and focus scores 0 because output splits evenly — about 33% each — across the three fields; publication-volume, citation-impact, and activity metrics are not yet available in this snapshot.
Collaboration pattern
Chen's cataloged work is thoroughly collaborative: all three 2026 papers list them as a middle author with no corresponding role, within a network of 14 distinct coauthors (collaboration score 54). The recurring co-author list — including Meiyi Qiang, Hengyi Feng, Hao Liang, Qihan Lin, Linzhuang Sun, Sizhe Qiu, Wentao Zhang, Hexiong Yang, Ran He, and Jie Cao — suggests one stable core group spanning the multimodal, post-training, and educational AI strands.
Research keywords
Highlights
A 591-run matched-seed null result on RLVR data policies
Training-free visual workspaces for sandboxed VLM agents
Capability-balanced SFT for K-12 educational foundation models
Timeline
- 2026Co-authored DataFlex-RL, a 591-run matched-seed evaluation platform finding that RLVR data policies add no reproducible gain over uniform GRPO.
- 2026Co-authored OmniEdu, applying capability-balanced SFT to lift Qwen 4-27B models across K-12 educational benchmarks.
- 2026Co-authored VLM-in-Sandbox, a training-free visual workspace for agentic visual reasoning in sandboxed vision-language models.
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 by Evolve from a limited snapshot; only 3 reviewed papers (all 2026 arXiv preprints) inform this profile.
- Output, impact, and activity dimensions are unscored — no publication counts, citation metrics, or multi-year history were available.
- 'Mingrui Chen' is a common Chinese name and no ORCID is provided; this profile is name-resolved from public bibliographic data and may merge or split identities.
- All cataloged papers are arXiv preprints; peer-reviewed venue status is unknown.