Evolve.Pub
Sign inScan your paper
CZ

Chen Zhang

Shanghai Jiao Tong University · China

Unclaimed · AI-structured profileOpenAlex

Builds memory, context, and orchestration infrastructure for LLM agents at Shanghai Jiao Tong University

5

papers here

3

reviewed by Evolve

0

first-author

0

corresponding

—

works · OpenAlex

1

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: insufficient dataOutputImpact: insufficient dataImpactFocus: 0/100FocusCollaboration: 68/100CollaborationActivity: 84/100ActivityReproducibility: 55/100Reproducibility

Output—

Not yet scorable: the record lacks the volume and venue history this axis needs — its only year-level count is 17 works in 2026. Output would normally measure publication volume, venue quality, and consistency of production over time.

Impact—

Not yet scorable: citation signals are effectively absent (h-index 1, i10-index 0), so no responsible influence estimate is possible. Impact would measure citation accumulation, uptake of the memory and context-assembly methods, and independent replication by other groups.

Focus0/100

Scores 0, mirroring an almost evenly split field distribution — LLM agent memory systems, LLM-driven data visualization (NLP/HCI), and AI agents for software engineering each take roughly one third. OpenAlex topic counts show the same spread: multimodal machine learning applications (6), topic modeling (5), AI-based problem solving and planning (3), advanced neural network applications (2), and NLP techniques (2). This reads as parallel workstreams rather than an established single specialty.

Collaboration68/100

Scores 68, drawn from 29 distinct co-authors on the OpenAlex record. A recurring core appears across papers — Haijie Xu, Peng Xu, Zuyu Zhang, Yuze Sun, Feng Tian, Long Wang and others (12 frequent collaborators listed). In the Evolve catalogue the authorship sits in the middle on three papers and last on two, never first or corresponding, consistent with substantive but shared contributions inside multi-person project teams.

Activity84/100

Scores 84, the strongest axis of this profile. The record counts 17 works in the recent three-year window, and its only year-level entry places all 17 in 2026. The Evolve catalogue corroborates the burst with five 2026 arXiv papers spanning agent memory, context assembly, data visualization, autonomous software development, and causal order recovery — a dense, current production cadence, with no earlier-year data on file.

Reproducibility55/100

Scores 55, a mean over 3 assessed papers. Per-paper readings are medium for MemForest, InsightChain, and Harness-of-Harness, and low for ContextPipe, whose evidence rests on just 3 SWE-bench Pro instances without quality metrics. Verdicts repeatedly cite missing significance tests, single-run experiments, and author-owned judge metrics — methods are described openly, but validation depth trails the output pace.

Research summary

Chen Zhang is a researcher at Shanghai Jiao Tong University whose recent work centers on the infrastructure of large language model agents — how they remember, assemble context, orchestrate each other, and build software autonomously. The OpenAlex-linked record (name-resolved from public bibliographic data; no ORCID provided) shows a young but fast-moving profile: 17 works in the recent three-year window, all attributed to 2026 in year-level counts, co-authored with 29 distinct collaborators. Research output splits almost evenly across three declared areas — LLM agent memory systems, LLM-driven data visualization at the NLP/HCI boundary, and AI agents for software engineering — each about a third of the field distribution, which yields a focus score of 0. Four of the five catalogued 2026 arXiv papers have been reviewed by Evolve: MemForest (event-tree memory management reporting 50% compression with 97–99.7% retention), InsightChain (a four-stage chain-of-insight visualization pipeline), Harness-of-Harness (planner–coder–QA loops for multi-day autonomous development), and last-authored ContextPipe (database-inspired context assembly). Evolve's recurring criticism is evaluation rigor: gains are frequently judge-scored, single-run, or unaccompanied by significance tests, producing a reproducibility mean of 55 over 3 assessed papers. With an h-index of 1 and an i10-index of 0, impact cannot yet be meaningfully scored.

Collaboration pattern

Collaboration is broad for an early-stage record: 29 distinct co-authors, with a stable recurring core (Haijie Xu, Peng Xu, Zuyu Zhang, Yuze Sun, Feng Tian, Long Wang, Junxi Wang, Te Sun, Jiayi Zhu, Siyuan Li, Xuyang Liu, Zichen Wen). Across the five catalogued 2026 papers the signature sits middle (three papers) or last (two), never first or corresponding — a shared-credit contributor pattern inside multi-person agent-systems teams.

Research keywords

LLM agentsagent memorymemory compressioncontext assemblyretrieval augmentationdata visualizationautomatic prompt optimizationLLM-as-judge evaluationautonomous software developmentmulti-agent orchestrationlatent causal order recovery

Highlights

Timeline

  1. 2026Last-authored ContextPipe, recasting LLM context assembly as database-style query execution for long-horizon agents.
  2. 2026Released Harness-of-Harness, wrapping coding agents in planner–coder–QA loops for multi-day autonomous development.
  3. 2026Published InsightChain, a four-stage LLM chain-of-insight pipeline for data visualization with vision-guided prompt optimization.
  4. 2026Published MemForest, an event-tree approach to agent memory management reporting 50% compression with 97–99.7% retention.
  5. 2026Co-authored Jigsaw-CRL on recovering global latent causal order from fragmented multi-client interventions; not yet reviewed in the Evolve catalogue.

Papers in the Evolve catalog

Frequent co-authors

Haijie Xu×1Peng Xu×1Zuyu Zhang×1Yuze Sun×1Feng Tian×1Long Wang×1Junxi Wang×1Te Sun×1Jiayi Zhu×1Siyuan Li×1Xuyang Liu×1Zichen Wen×1

Is this you? Claim this profile

Enter the email you published with. We'll send a one-time claim link; after claiming you can edit the introduction or hide this page. Requests from a mismatched address go to manual review.

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 come from OpenAlex only (identity: openalex); no ORCID is linked, and 'Chen Zhang' is a common name, so this profile is name-resolved from public bibliographic data and may merge or split homonyms.
  • Citation coverage is minimal — h-index 1 and i10-index 0 — so the output and impact axes are left unscored.
  • Year-level publication counts cover 2026 only (17 works); no earlier career history is available in the data.
  • All catalogued papers are arXiv preprints; Jigsaw-CRL additionally lacks an Evolve review, keywords, and a reproducibility reading.
Chen Zhang — Researcher Profile | Evolve.Pub