Jingbo Yang

Bio

Jingbo Yang is a third-year Ph.D. student in computer science at the University of California, Santa Barbara (UCSB), advised by Prof. Shiyu Chang. His research focuses on building efficient and capable LLM agents, and on evaluating them reliably in realistic settings, spanning the following directions:

• Efficient LLMs and Agents: reducing the inference cost of LLMs and agents without sacrificing performance, spanning KV cache reuse, adaptive reasoning effort selection, and agent memory compaction.
  Works: KVLink (NeurIPS'25), Ares (NeurIPS'26), ReCAP (preprint)
• Agent Planning and Self-Evolution: enabling agents to solve complex, long-horizon tasks through dynamic planning, and to continually improve by evolving reusable skills and agent harnesses.
  Works: WebDART (ACL'26 Findings), FederatedSkill (preprint), ScholarEvolve (preprint)
• Evaluation of LLM and Multimodal Agents: building benchmarks that test agents and LLM judges in realistic scenarios, such as memory in multi-party conversations, compliance checking in dialogue systems, and multimodal agents auditing interactive 3D worlds.
  Works: GroupMemBench (NeurIPS'26 E&D Track), CompliBench (preprint), WorldAuditBench (preprint)

Before joining UCSB, he was a master's student in the Data Lab at Rice University under the supervision of Prof. Xia (Ben) Hu, and received a bachelor's degree in computer science from Wuhan University under the supervision of Prof. Yaochu Jin.

News

• Oct 2026Released TIPER, which uncovers robustness weaknesses in LLM agents by evolving user personas under task-integrity constraints.
• Sep 2026Released three new papers: ScholarEvolve on research-guided agent harness evolution, ReCAP on memory compaction for LLM agents, and WorldAuditBench on 3D world auditing with multimodal agents.
• Sep 2026Ares is accepted to NeurIPS 2026, and GroupMemBench is accepted to the NeurIPS 2026 Evaluations & Datasets Track.
• Jun 2026Started a research internship at Microsoft in Redmond, WA.
• Jun 2026Released FederatedSkill, a federated learning framework for agentic skill evolution.
• Apr 2026WebDART is accepted to Findings of ACL 2026.
• Apr 2026Released CompliBench, a benchmark for LLM judges on compliance violation detection.
• Sep 2025KVLink is accepted to NeurIPS 2025.

Publications

^‡ indicates authors with equal contribution/advising.

Peer-Reviewed

Ares: Adaptive Reasoning Effort Selection for Efficient LLM Agents

Jingbo Yang, Bairu Hou, Wei Wei, Yujia Bao^‡, Shiyu Chang^‡

NeurIPS 2026 Advances in Neural Information Processing Systems

Paper Code

GroupMemBench: Benchmarking LLM Agent Memory in Multi-Party Conversations

Jingbo Yang^‡, Kwei-Herng Lai^‡, Xiaowen Wang, Shiyu Chang, Yaar Harari, Evgeniy Gabrilovich

NeurIPS 2026 Advances in Neural Information Processing Systems, Evaluations & Datasets Track

Paper Code

WebDART: Dynamic Decomposition and Re-planning for Complex Web Tasks

Jingbo Yang, Bairu Hou, Wei Wei, Shiyu Chang, Yujia Bao

ACL 2026 Findings Findings of the Association for Computational Linguistics

Paper Code

KVLINK: Accelerating Large Language Models via Efficient KV Cache Reuse

Jingbo Yang^‡, Bairu Hou^‡, Wei Wei, Yujia Bao, Shiyu Chang

NeurIPS 2025 Advances in Neural Information Processing Systems

Paper Code

Self-Resource Allocation in Multi-Agent LLM Systems

Alfonso Amayuelas, Jingbo Yang, Saaket Agashe, Ashwin Nagarajan, Antonis Antoniades, Xin Eric Wang, William Wang

ICML 2025 Workshop Multi-Agent Systems (MAS) Workshop

Paper

Preprints

Uncovering Robustness Weaknesses in LLM Agents via Persona Evolution with Task-Integrity Constraints

Yutao Xie, Jingbo Yang, Guanyu Yao, Xiaoyu Yang, Duo Ding, Shiyu Chang

Preprint arXiv 2610.16155

Paper Code

Learning from Research: Toward Lifelong Agent Harness Evolution

Jingbo Yang, Kwei-Herng Lai, Xiaowen Wang, Yaar Harari, Evgeniy Gabrilovich, Shiyu Chang

Preprint arXiv 2609.40169

Paper Code

Persistent Context Graphs for Efficient Memory Compaction in LLM Agents

Jingbo Yang, Kwei-Herng Lai, Xiaowen Wang, Zhaoxuan Tan, Pei Zhou, Mengting Wan, Yaar Harari, Evgeniy Gabrilovich, Shiyu Chang

Preprint arXiv 2609.40118

Paper Code

WorldAuditBench: Interactive 3D World Auditing with Multimodal Agents

Ziyan Jiang^‡, Jingbo Yang^‡, Jiabao Ji^‡, Yujian Liu, Qiucheng Wu, Tommi Jaakkola, Yang Zhang, Shiyu Chang

Preprint arXiv 2609.40325

Paper Project

FederatedSkill: Federated Learning for Agentic Skill Evolution

Jingbo Yang^‡, Guanyu Yao^‡, Yang Zhang, Ramana Rao Kompella, Gaowen Liu, Shiyu Chang

Preprint arXiv 2606.03143

Paper Code

CompliBench: Benchmarking LLM Judges for Compliance Violation Detection in Dialogue Systems

Jingbo Yang^‡, Guanyu Yao^‡, Bairu Hou, Xinghan Yang, Nikolai Glushnev, Iwona Bialynicka-Birula, Duo Ding, Shiyu Chang

Preprint arXiv 2604.12312

Paper Code

Vitæ

• Microsoft 2026 Summer

  Research Intern
  Redmond, WA
  
• Cresta AI 2025-2026

  Research Intern
  Remote
  
• Accenture 2025 Summer

  Research Intern
  Mountain View, CA
  Mentor: Yujia Bao
• UC Santa Barbara 2024 - now

  Ph.D. Student
  Santa Barbara, CA
• AutoEdge.ai 2023-2024

  Machine Learning Engineer Intern
  Sacramento, CA
• SAP Labs 2023 Summer

  Software Engineer Intern
• Rice University 2022 - 2023

  Master student
  Houston, TX
• Wuhan University 2018 - 2022

  B.S. student
  Wuhan, China

Misc

This website was built with jekyll based on a template by Martin Saveski.