My name is Jinxiang Xie (pronounced “JIN-shee-ahng SHEE-eh”). I am an M.Sc. student at Nanjing University, supervised by Prof. Hao Dong. Previously, I was a research intern at Microsoft, supervised by Principal Researcher Justin Ding, and at Peking University, supervised by Prof. Xiaojun Wan.
My research explores how large language models can address complex problems, with a focus on harness engineering, agentic reinforcement learning, and collective intelligence.
Feel free to reach out if you’d like to discuss research or explore potential collaboration!
AI Researcher
- Research focus on LLMs and NLP
- Internships at top institutions
- Publications at AAAI and ACL
Writer & Builder
- Technical blogs with 500K+ views
- Active on Xiaohongshu
- Articles about Tech & Humanities
Life Explorer
- Visited 9 countries worldwide
- Traveled to 32 provinces in China
- Rich experience in social work
Publications
arXiv 2026
Semantic Atoms for Scalable Text Analysis in Data Analytics in the Era of LLM-Based Agents
ACL 2026 Findings
CAST: Achieving Stable LLM-based Text Analysis for Data Analytics
Education
2025.09 — PresentNanjing UniversityMaster of Science at Kuang Yaming Honors School
2021.09 — 2025.06Beijing Jiaotong UniversityBachelor of Science in Information and Computing Science
Internships
2026.08 — PresentMind Lab (Mindverse)Algorithm Intern, Continual Learning Group
2024.08 — 2025.08MicrosoftResearch Intern, Data, Knowledge and Intelligence (DKI) Group
2023.11 — 2024.08Peking UniversityResearch Intern, Wangxuan Institute of Computer Technology
Open Source
Skills for AI Agents
Scientific Research Skills
Not another tool list — each skill encodes a complete research workflow: literature search, paper reading, social-media triage, related-work survey, Zotero management, and figure generation.
Research Tool
Auto Paper Harvester
Routes through publisher TDM APIs (Wiley/Elsevier/Springer) → OA fallbacks (Crossref/OpenAlex/Unpaywall) → optional Playwright browser fallback. Ships as both a CLI and an agent skill.
Blogs
Oct 2026
中 / EN
One Memory, Many Agents: A Shared Memory Layer with Per-Agent Recall
Personal and coding agents each keep their own memory. I built one shared Markdown memory with per-agent recall profiles and an automatic write path, then evaluated...
Sep 2026
中 / EN
Where Collective Intelligence Hits Its Ceiling
A closer look at orchestration, multi-agent reinforcement learning, and an offline voting audit: why diverse agents do not automatically deliver collective gains, and what it takes...
Aug 2026
Consolidating an Agent That Outgrew Itself
After months of heavy use, my AI agent had accumulated 272 skills, plaintext API keys, and a 43K-token fixed context — with no way to check...
Jun 2026
Staying Grounded in Fast-Paced AI Research
Three principles for navigating the AI research firehose: prioritizing information sources (blogs over papers for frontier insights), building domain-specific harnesses instead of chasing hype, and constructing...
Jan 2026
My 2025 | Through Romance, Finding Stability
From adventure and exploration in the first half to stability and reflection in the second half - sharing thoughts on compound interest of time, recording in...
Nov 2025
Claude Code Skills and Subagents in Practice
Two production-grade systems for real research scenarios: a paywall-crossing paper harvester and a self-iterating AI Scientist, demonstrating how Skills and Subagents upgrade LLMs into research infrastructure....
Jun 2025
The Limits of My Language Mean the Limits of My World
Drawing on Wittgenstein's philosophy, a recent paper argues that our existing language might be the fundamental bottleneck.
Feb 2025
As the World Reorders: Interviews, Content Creation, and Digital Nomadism
From interview reflections and Coffee Chats to AI-assisted writing and a digital-nomad experiment in Bali, this essay explores how to build an inner anchor in an...
Jan 2025
Beyond the Future of AI: The Dreams and Deceptions of Cryptocurrency
From Web 3.0 to Bitcoin breaking $100K, from speculators to true believers - exploring the technical potential of cryptocurrency, its future direction, and the possibility of...