Early-career researcher spanning LLM agent evaluation, personalized language models, and AI-driven cognitive diagnosis
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
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—
Insufficient data to score: only an h-index of 1 and i10-index of 0 exist in the OpenAlex snapshot. This axis would measure total publication volume and per-year productivity across venues once a fuller, verified publication record is available.
Impact—
Insufficient data: citation counts are absent beyond an h-index of 1, so influence cannot be scored yet. This axis would measure citations, citation trajectory, and uptake of the 2026 preprints as they accumulate.
Focus0/100
The focus score is 0, the lowest band, and the raw data supports it: the field distribution splits exactly into thirds — LLM agent evaluation, personalized LLMs, and AI for education/cognitive diagnosis (0.333 each) — while topics scatter across topic modeling (10), NLP techniques (6), multimodal machine learning (6), adversarial robustness (4), and explainable AI (3). This is a broad, exploratory portfolio rather than a concentrated agenda.
Collaboration51/100
Collaboration scores 51, mid-range. The record shows 15 works over the last three years with 12 distinct coauthors; recurring partners include Minghang Liu, Huawei Shen, Xueqi Cheng, Zhichao Shi, and Xuhui Jiang. On the three cataloged papers Wang is last author once and a middle author twice, none corresponding — consistent with contributing within team-led projects rather than yet leading an independent group.
Activity81/100
Activity scores 81, the strongest axis: counts by year show 13 works in 2026 against 1 each in 2022, 2024, and 2025, and 15 works across the recent three-year window. All three Evolve-cataloged papers appeared on arXiv in 2026, indicating an intense current output phase rather than a sustained long-term cadence.
Reproducibility55/100
Reproducibility scores 55, based on Evolve's assessment of 3 papers (reproMean 55). All three carry medium reproducibility readiness; reviews note methodological care — e.g., evaluation across all 24 LaMP cells in the personalization work — but also self-defined metrics, potential leakage, and no code release stated for the personalization paper, so practical reproducibility remains untested.
Research summary
Yuanzhuo Wang is an early-career researcher at the Institute of Computing Technology (China) whose OpenAlex-resolved record shows output concentrated in 2026: 13 works that year against 1 each in 2022, 2024, and 2025, with an h-index of 1 and an i10-index of 0, so citation impact has not yet accumulated. Three 2026 arXiv preprints anchor the Evolve catalog across three co-equal fields (0.333 each): LLM agent evaluation (DynSTEER, stage-wise milestone-graph trajectory evaluation with early stopping; Wang is last author), cognitive diagnosis (PLCD, replacing ID embeddings with LLM-derived process graphs and response-calibrated memory), and personalized LLMs (ENOUGH, minimal-sufficient user profiles, best across all 24 LaMP cells). Evolve's reviews judge the designs sound-to-rigorous while flagging self-defined metrics, potential data leakage, LLM cost fairness, and unstated code release; all three papers carry medium reproducibility readiness, averaging 55 on the reproducibility axis. Topic signatures — topic modeling (10), NLP techniques (6), multimodal machine learning (6), adversarial robustness (4), explainable AI (3) — reinforce this breadth, consistent with a focus score of 0 on an evenly split field distribution. Collaboration is mid-range (51): 15 works over the last three years with 12 distinct coauthors.
Collaboration pattern
Wang works in team settings: 12 distinct coauthors over the last three years alongside 15 works, with repeat partners including Minghang Liu, Huawei Shen, Xueqi Cheng, Zhichao Shi, Xuhui Jiang, Wenjie Zhang, Chengjin Xu, and Jian Guo. Authorship on the three cataloged 2026 papers is last author once (DynSTEER) and middle author twice, with no corresponding-author role — a mid-range collaboration score (51) consistent with contributing within larger groups rather than yet leading an independent lab.
Research keywords
Highlights
Stage-wise milestone-graph evaluation for LLM agents
Process-grounded modeling replaces ID embeddings in cognitive diagnosis
Minimal sufficient user profiles for personalized LLMs
Timeline
- 2026DynSTEER: dynamic stage-wise trajectory evaluation for LLM agents posted to arXiv, with Wang as last author.
- 2026Beyond ID Embeddings (PLCD) applies process-grounded language modeling to cognitive diagnosis, with Wang as a middle author.
- 2026Less Is Personal (ENOUGH) learns minimal sufficient user profiles for personalized language models, with Wang as a middle author.
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 data only; other databases may differ.
- Citation impact metrics are thin (h-index 1, i10-index 0) because output is concentrated in 2026; impact needs time to materialize.
- All three cataloged papers are 2026 arXiv preprints, not yet verified as peer-reviewed publications.
- "Yuanzhuo Wang" carries common-name ambiguity; this profile is name-resolved from public bibliographic data (OpenAlex identity, ORCID 0000-0003-4343-1233).