Xing Sun: last-author researcher on LLM-agent long-term memory and graph foundation models (two 2026 arXiv papers).
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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—
Not yet scored: the catalog lacks full publication-count and venue data for this name, so research output cannot be measured. This axis would aggregate total papers, venue tiers, and first/corresponding-author volume over time.
Impact—
No citation or influence data is available yet, so impact is unscored. Once linked, this axis would track citations, benchmark adoption, and downstream reuse of the latent-memory and graph foundation model methods.
Focus0/100
Scores 0/100 on this radar, the lowest band: the reviewed output splits evenly (0.5/0.5) between NLP work on LLM-agent long-term memory and personalization and machine-learning work on graph representation learning and retrieval-augmented generation. The two threads are methodologically linked — both papers use graph neural networks and RAG — but the even split reads as an exploratory, dual-track agenda rather than a tightly concentrated niche.
Collaboration63/100
Scores 56/100. The record shows 16 distinct coauthors across the reviewed work, with 12 recurring partners including Siyu An, Junnan Dong, Ruizhi Qiao, Linhao Luo, and Ruifeng Xu. On both 2026 arXiv papers Sun sits in the last author position without corresponding status, a pattern consistent with a supervisory or senior-contributor role within a moderately sized, stable collaboration network rather than a solo or narrowly paired style.
Activity75/100
Insufficient data: only the 2026 arXiv entries are on record, so yearly productivity cannot be scored. This axis would measure publication cadence and sustained research activity across recent years.
Reproducibility32/100
Scores 20/100, computed as the mean reproducibility readiness across the 2 reviewed papers, both rated low. Evolve's verdicts cite concrete gaps: the LGM memory paper lacks significance tests, a stated judge, and released code, while the WFM paper's 10.5x speedup and SOTA claims are weakened by absent statistics and a leaked 'planning values' note. Independent replication is currently difficult for both lines.
Research summary
Xing Sun is an early-stage profile in the Evolve catalog, name-resolved from public bibliographic data and currently documented by two 2026 arXiv papers on which Sun appears as last author, non-corresponding. The reviewed work splits evenly (0.5/0.5) between two adjacent areas: natural language processing on LLM-agent long-term memory and personalization, and machine learning on graph representation learning and retrieval-augmented generation. 'Disentangling Long-Term Memory via Latent Neuro-Symbolic Reasoning' introduces LGM's query-conditioned latent graphs, which show large gains on personalization benchmarks. 'WFM: Wiki Foundation Model for Complex Agentic Reasoning' builds a graph foundation model over dense LLM-Wiki graphs, claiming a 10.5x speedup and state-of-the-art question-answering gains. Together the two papers trace one agenda: graph-structured, retrieval-grounded memory for LLM agents. Evolve's reviews credit the ambition in both but rate reproducibility readiness low: missing significance tests, no stated judge, no released code, and, for WFM, a leaked 'planning values' note. The collaboration record lists 16 distinct coauthors with 12 recurring partners. Output, impact, and activity metrics are not yet linked, so overall productivity and citation influence cannot be assessed from this snapshot.
Collaboration pattern
The footprint centers on a stable mid-size network: 16 distinct coauthors, with 12 recurring partners including Siyu An, Junnan Dong, Ruizhi Qiao, Linhao Luo, Ruifeng Xu, Xiaoyu Shen, Keyu Chen, and Xinghao Chen. Sun's consistent last-author, non-corresponding position on both 2026 papers is a signature consistent with a senior or supervisory role rather than first-hand execution.
Research keywords
Highlights
Query-conditioned latent graphs for long-term agent memory
Wiki Foundation Model claiming a 10.5x speedup
Consistent last-author role across both 2026 papers
On both reviewed arXiv papers Sun is last author without corresponding status, working within a network of 16 distinct coauthors and 12 recurring partners; both carry low reproducibility readiness, making code and statistics release the highest-leverage next step.
Timeline
- 2026'WFM: Wiki Foundation Model for Complex Agentic Reasoning' appears on arXiv with Sun as last author, claiming a 10.5x speedup and SOTA QA gains; Evolve rates reproducibility readiness low.
- 2026'Disentangling Long-Term Memory via Latent Neuro-Symbolic Reasoning' appears on arXiv; LGM's query-conditioned latent graphs report large personalization gains with low reproducibility readiness.
- 2026Both papers enter the Evolve review catalog, spanning NLP agent memory and graph representation learning, with 16 distinct coauthors on record.
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 public bibliographic data; output, impact, and activity axes are missing because the underlying counts are not yet linked.
- 'Xing Sun' is a common name; this profile is name-resolved from public bibliographic data, and institution, country, and ORCID are unverified — the Chinese-character spelling of the name is not confirmed.
- The reproducibility score of 20 rests on a small sample: only 2 arXiv papers, both from 2026, were reviewed.
- No career-history information is available in the input data.