Stanford researcher spanning LLM-agent evaluation infrastructure and privacy-preserving synthetic sensor data
3
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
26
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—
Publication-output data for this axis is not yet available in the underlying record, so no score is given. This dimension would measure total scholarly output — number of works, career-long output trajectory, and distribution across venues and roles.
Impact—
Impact signals (citations and uptake measures) are not yet available for this profile, so the axis is left unscored. It would measure how far the work propagates — citation accumulation, influence on follow-up research, and reach within and beyond the field.
Focus0/100
The focus score of 0 reflects an unusually broad record: field distribution splits exactly 50/50 between IoT and ubiquitous computing (sensor privacy) and machine learning (LLM agent evaluation), and the leading OpenAlex topics span at least five areas — topic modeling (35), privacy-preserving technologies (20), stock market forecasting (15), multimodal machine learning (15), and neural network applications (14). This is a multi-line portfolio rather than a single narrowly defined research thread.
Collaboration84/100
A collaboration score of 84 rests on measurable breadth: 67 distinct coauthors within the recent three-year window alongside 99 works, indicating consistently team-based research. Twelve frequent collaborators — including Yuandong Pan, Jie Wang, Michael D. Lepech, Bradford Campbell, and Lin Shi — recur across the record, and all three 2026 papers in the Evolve catalog are multi-authored, with PrivateHub the only first-author position among them.
Activity100/100
Activity is scored at the ceiling of 100 on a steep trajectory: recorded works climb from 1 in 2010 and 3 in 2015 through 14 in 2021 and 21 in 2022 to 45 in 2025 and 40 already in 2026, with 99 works in the most recent three-year window. The 2021-2026 curve indicates a sustained and still-accelerating publication pace concentrated in the last five years.
Reproducibility78/100
Reproducibility averages 78 across the two papers Evolve scored (reproMean 78, reproPapers 2). The spread is visible in the catalog: the Harbor infrastructure paper carries high reproducibility readiness as openly released tooling, while PrivateHub (synthetic sensor data) and EGT-KG (retrieval-augmented QA) sit at medium — the latter criticized for reporting gains without significance tests or released code. Open infrastructure work anchors the score; per-paper documentation gaps hold it below the top.
Research summary
Jiechao Gao is a researcher at Stanford University whose recorded work spans two roughly equal lines: evaluation infrastructure for LLM agents, and privacy-preserving generation of sensor-intensive IoT data. The OpenAlex record lists an h-index of 26 and an i10-index of 39, with 99 works in the most recent three-year window and 67 distinct coauthors. Output has accelerated sharply: 14 recorded works in 2021 and 21 in 2022 rise to 45 in 2025 and 40 in 2026, underpinning a maximal activity score. The field distribution splits evenly between IoT and ubiquitous computing (sensor privacy) and machine learning (LLM agent evaluation), with topic frequency led by topic modeling (35) and privacy-preserving technologies (20). Three 2026 arXiv papers reviewed by Evolve show the range: Harbor Adapters and Harbor-Index unifies 80+ agentic benchmarks in one harness with an 82-task index on which the top score is 28%; first-authored PrivateHub applies contrastive diffusion to mask private activities in shared sensor streams; EGT-KG tests typed knowledge-graph retrieval for scientific QA with small language models. Across them the posture is infrastructure- and systems-oriented, with an average reproducibility of 78 across the two papers scored.
Collaboration pattern
The record is strongly collaborative: 67 distinct coauthors over the last three years against 99 works, with a collaboration score of 84. Twelve collaborators recur, including Yuandong Pan, Jie Wang, Michael D. Lepech, Bradford Campbell, Lin Shi, Haowei Lin, Zixuan Zhu, Xiaoyue Zhou, Xiang Li, Xiangning Lin, Yaxuan Deng, and Han Xu. The 2026 catalog papers show team-based middle-author roles plus one first-author lead (PrivateHub), pointing to contribution through shared systems work.
Research keywords
Highlights
Unified harness for large-scale agentic evaluation
First-authored diffusion approach to sensor privacy
Typed knowledge-graph retrieval for small language models
Timeline
- 2026Co-authored EGT-KG, proposing evidence-grounded typed knowledge-graph retrieval for scientific question answering with small language models.
- 2026First-authored PrivateHub, a contrastive-fine-tuned diffusion model that masks private activities in shared smart-home sensor streams.
- 2026Co-authored Harbor Adapters and Harbor-Index, unifying 80+ agentic benchmarks under one harness and releasing an 82-task index.
Papers in the Evolve catalog
Harbor Adapters and Harbor-Index: Infrastructure and a Curated Meta-Dataset for Large-Scale Agentic Evaluation
- first author
PrivateHub: Contrastive Diffusion Model for Private Sensor-Intensive Environment Data Generation
EGT-KG: Evidence-Grounded Typed KG Retrieval for Practical Scientific QA with Small Language Models
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.
- Bibliometrics are computed from OpenAlex only; the output and impact dimensions lack sufficient underlying data and are not scored.
- Identity is name-resolved from public bibliographic data (OpenAlex record, ORCID 0000-0003-0628-1416); common-name collisions cannot be fully excluded.
- Evolve verdicts cover three 2026 arXiv papers; the broader publication corpus has not been reviewed here.
- The reproducibility score draws on only two scored papers (reproPapers 2) and may not generalize across the full record.