Luxin Zhang

Research · Official Researcher Profile

Luxin Zhang

Researcher in AI Systems

I study how AI agents learn from experience and interact with the systems around them. My work focuses on reliable memory for lifelong agents, documentation and interfaces for agent programming, and AI tools for scientific discovery and health. I connect these questions with my background in software engineering, secure systems, and computer vision.

Research Interests

Lifelong Agent Memory

How agents turn individual experiences into reliable shared procedures, with scope-aware promotion, evidence gates, and resistance to memory poisoning.

Agent-Facing Interfaces

How documentation, interface design, and tool access affect code generation and correctness, including controlled studies of StableHLO.

AI for Scientific Discovery & Health

Evidence mapping for scientific research and digital phenotyping for neurodegenerative disease detection.

Accepted Papers

Accepted poster · 3rd Workshop on Lifelong Agents (AIAS 2026)

From Episodes to Procedures: Scope-Aware Memory Promotion for Lifelong Enterprise Agents

Luxin Zhang · November 7, 2026 · DOI: 10.5281/zenodo.24063133

ScopeAware turns successful agent experiences into shared procedures only when the evidence is reliable and the scope is explicit. Across enterprise support and scientific-facility evaluations, it improves resolution accuracy while sharply reducing harmful memory promotion.

View paper details →

Accepted poster · VISA 2026 at IEEE/ACM MICRO 2026

What Makes a Virtual ISA Agent-Programmable? A Controlled Study of Agent-Facing Documentation for StableHLO

Luxin Zhang · November 2026 · DOI: 10.5281/zenodo.24062377

A controlled study of how documentation format and delivery affect AI agents programming StableHLO. Syntax alone produces large gains, while matched prose outperforms JSON contracts and upfront documentation consistently beats on-demand retrieval.

View paper details →

Research Writing

Research Paper

Hypothesis Atlas: An Open-Source Evidence Mapping Platform for Multi-Domain Scientific Research

Hypothesis Atlas introduces an open-source, hypothesis-first evidence mapping workflow that retrieves live papers, datasets, and early signals, then surfaces disagreements, gaps, and research opportunities through epistemically-labeled synthesis.

Read More →

Book Chapter

NEURORA: Mandala-Based Digital Phenotyping for Early Detection of Parkinson's Disease and Alzheimer's Disease in the Era of Artificial Intelligence

A book chapter exploring digital phenotyping approaches using mandala-based assessments for early detection of neurodegenerative diseases, published in 'Where Art and Technology Resonate: Binnovative Approaches to Digital Health and Wellbeing'.

Read More →

Research Software

Hypothesis ATLAS

AI-powered evidence mapping system for biotech research that discovers, organizes, and visualizes signals from papers, datasets, and scientific updates.

View Repository →

Academic Service

Reviewer

AgentEvalSci 2026 — Fall NYC

EvalScience

November 2026 · New York, New York

Official website ↗

Reviewer

Third Workshop on Synthetic & Adversarial Forensics (SAFE 2027)

SAFE Workshop

November 2026 · Montréal, Québec

Official website ↗

Program Committee Member (Reviewer)

AAAI-27 Undergraduate Consortium

Association for the Advancement of Artificial Intelligence

October 8–14, 2026 · Palo Alto, California

Official website ↗

Reviewer

2027 ASPPH Annual Meeting for Academic Public Health

Association of Schools and Programs of Public Health

October 2026 · Washington, District of Columbia

Official website ↗

Reviewer

EMNLP 2026 Industry Track

Association for Computational Linguistics

July 22–24, 2026 · Stroudsburg, Pennsylvania

Official website ↗
View ORCID Record ↗View All Judging Experience →