1. 3DCodeEdit: Benchmarking Instruction-Guided 3D Program Editing
In submission
Xinran Zhao
PhD Candidate · Carnegie Mellon University
I am a fourth-year PhD candidate at CMU LTI, advised by Sherry Wu. My PhD is supported by the Amazon AI PhD Fellowship.
At Google DeepMind, I worked as a student researcher on LLM planning with Azade Nova and Hanie Sedghi (2024), and on long-horizon code generation and continual learning with Azade (2026). At Ai2 (2025), I worked on deep research agents with Varsha, Aakanksha, Joseph, Jay, and Jena.
I received my MSCS from Stanford, where I worked with Christopher Manning, and my bachelor's degree from HKUST. At HKUST and during my exchange at Cornell, I was fortunate to work with Yangqiu Song, Dit-Yan Yeung, and Claire Cardie.
I always feel lucky to meet great mentors along my journey, including but not limited to, Shikhar Murty, Hongming Zhang, and Esin Durmus. I serve as a general chair of FrontierIR @ AAAI 2027 and as an area chair for ACL, EMNLP, and NAACL.
I am on the 2026–27 industry job market, with a flexible graduation date!
Research
I am interested in meaningful research on natural language processing (NLP) and language model agents. My goal is to build efficient, effective, and faithful systems that work with people on complex real-world tasks. Recently, I focus on information seeking and planning in long-horizon scenarios, organized around three questions:
Information Seeking & Delivery
How can agents proactively help humans seek information, learn, and stay educated amid massive artificial artifacts?
Escalation (proactive delivery, in submission) · DR Tulu (ICML'26) · Intent-aware QA (ICLR'26) · Revela (ICLR'26) · Mixture of Retrievers (EMNLP'25) · PIR (COLM'24) · Dense X (EMNLP'24) · MixGR (EMNLP'24) · Thrust (when to retrieve, NeurIPS'23)
Reasoning, Planning, Forecasting & Coding Agents
How can agents learn, maintain memory, and evolve over complex long-horizon tasks?
3D Code Editing (In submission) · Tea Leaves (In submission) · ReSHAPE (BiAlign@NeurIPS'26) · Harness Adaptation (In submission) · Pivot-ICL (In submission) · Ramon Llull (LM4Sci@COLM'25) · cAST (EMNLP'25) · Action Sequence Similarity (ICLR'25)
Persuasion & Human–AI Alignment
How can models stay aligned with human motives, goals, and intents under high automation, including through multi-turn behaviors and how multimodal inputs shape model representations?
Real-Time Debate (In submission) · Strategic Debate (ICLR'26) · SPHERE (ACL'25) · FaR (confidence calibration, ACL'24) · Argument Persuasion (ACL'21) · Probing Toxic Content (ACL'21)
Selected Honors and Awards
- Best Paper Award, FrontierIR Workshop at AAAI 2026 (Revela)
- Amazon AI PhD Fellowship, 2025 (two years of full tuition and stipend; $150,000 in AWS credits)
- Google RSGCP Award on LLM Planning ($30,000), 2024
- Google RS Conference Travel Scholarship ($3,000), 2024
- HKUST Academic Achievement Medal (Top 1%), 2021
Selected Research
All research →Selected publications and ongoing work. * Equal contribution.
2. Planning while Listening and Speaking for Real-Time Competitive Debate
In submission
3. How Should Planning Agents ReSHAPE Their Search Space?
BiAlign@NeurIPS'26
4. Reading the Tea Leaves, Again: Can Large Language Models Predict Language Model Training Runs?
In submission
5. Knowing When to Stop: The Evidence–Action Gap in Coding Agents
In submission
6. Better Harnesses, Smaller Models: Building 90% Cheaper Agents via Automated Harness Adaptation
In submission
7. Improving Attributed Long-form Question Answering with Intent Awareness
ICLR'26
8. Strategic Planning and Rationalizing on Trees Make LLMs Better Debaters
ICLR'26
9. The Ramon Llull's Thinking Machine for Automated Ideation
LM4Sci@COLM'25
10. MoR: Better Handling Diverse Queries with a Mixture of Sparse, Dense, and Human Retrievers
EMNLP'25 [Code]
11. cAST: Enhancing Code Retrieval-Augmented Generation with Structural Chunking via Abstract Syntax Tree
EMNLP'25 [Code]
12. Improving Large Language Model Planning with Action Sequence Similarity
ICLR'25
13. Beyond Relevance: Evaluate and Improve Retrievers on Perspective Awareness
COLM'24
14. Fact-and-Reflection (FaR) Improves Confidence Calibration of Large Language Models
ACL'24
15. Thrust: Adaptively Propels Large Language Models with External Knowledge
NeurIPS'23
Large-scale Collaborations
Academic Events
- Winter, 2027 General chair, FrontierIR, the Second Workshop on New Frontiers in Information Retrieval at AAAI 2027.
- Fall, 2026 Invited panelist, Global Embodied Ecosystem Night at IROS 2026.
- Summer, 2026 Invited talk at Google DeepMind: Continual Learning for Coding Agents.
- Summer, 2025 Invited talk at UW H2 Group: Improving Large Language Model Planning with Action Sequence Similarity.
- Summer, 2024 Invited talk at Samaya AI: Towards Building Retrieval Systems in Complex Realistic Scenarios.
- Summer, 2022 Invited talk at the ICML Commonsense Tutorial: On Task Difficulty of Few-shot Learning.
- Spring, 2022 Invited talk at NUS, Department of Marketing: NLP as a Tool for Scientific Discovery.
Service
- General chair, FrontierIR, the Second Workshop on New Frontiers in Information Retrieval at AAAI 2027.
- Area chair for ACL and EMNLP (2024–present), and NAACL (2025–present).
- Reviewer for NeurIPS (2024–present), ICLR (2025–present), ICML (2025–present), and COLM (2024–present); previously for ACL, EMNLP, NAACL, EACL, AACL, AAAI, and ECCV.
Miscellaneous
Besides research, I also enjoy:
- Bodybuilding: I can deadlift 140 kg and squat 140 kg for sets.
- Debate: I was a member of the HKUST Mandarin Debate Team. We won first prize at the Bay Area Debate Championship and placed in the top eight at the sixth and seventh International Mandarin Debate Invitational in Singapore.
- Poetry: I write traditional Chinese poetry and have published a few poems in newspapers.
- Archery: I was an active member of the HKUST Archery Club.