Xiaoxuan (Mandy) Wang
Final-year Ph.D. Student
xw27 [at] cs.ucla.edu
About Me
I am a final-year Ph.D. student in Computer Science at the University of California, Los Angeles (UCLA), advised by Prof. Wei Wang and collaborating closely with Prof. Yizhou Sun.
My research focuses on agentic training and coding agents, spanning three complementary components: post-training agents for interactive tasks, building harnesses for long-horizon control, and verifying environments, rewards, and trajectories. I am also interested in LLM reasoning, including self-correction, tool use, and mathematical and scientific problem-solving.
I organize the UCLA Data Mining reading group. Before UCLA, I earned my B.S. in Computer Science from the University of Illinois Urbana-Champaign in May 2022.
I am looking for full-time Research Scientist positions in agentic post-training and coding agents, starting in late 2026.
News
- Sep 2026 Released ACLArena, a study of agent continual learning across multi-stage post-training, together with code and a model collection.
- Jun 2026 HarnessBridge is accepted to EMNLP 2026 Findings.
- Jun 2026 Released HarnessBridge on learnable control for long-horizon coding agents.
- Mar 2026 Invited talk on ARLArena at NICE (Nexus for IntelligenCE).
Publications
(* denotes equal contribution)
Agentic RL & LLM Agents
- Preprint 2026 ACLArena: Agent Continual Learning in Multi-stage Post-training [Paper] [Code] [Models]
- ICML 2026 ARLArena: A Unified Framework for Stable Agentic Reinforcement Learning [Paper] [Code] [Model]
- Preprint 2026 When Vulnerability Is Not Exploitation: Reward Hacking in Coding Agents
- EMNLP Findings 2026 HarnessBridge: Learnable Bidirectional Controller for LLM Agent Harness [Paper] [Code] [Models]
LLM Reasoning, Post-Training & Evaluation
- Preprint 2025 From Solving to Verifying: A Unified Objective for Robust Reasoning in LLMs [Paper]
- NeurIPS MathAI 2025 EAST: Entropy-Based Adaptive Weighting for Self-Training [Paper] [Code]
- ICML 2024 SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models Media coverage: Nature News Feature [Paper] [Code] [Website]
- KDD 2026 MatSciBench: Benchmarking the Reasoning Ability of Large Language Models in Materials Science [Paper] [Code] [Dataset]
Clinical & Domain Applications
- Preprint 2024 CliBench: Multifaceted Evaluation of LLMs in Clinical Decision Making [Paper] [Code] [Website]
- AAAI 2025 Memorize and Rank: Evaluating LLMs for Clinical Diagnosis Prediction Also at NeurIPS GenAI4Health Workshop 2024 [Paper]
Earlier Work (NLP & Speech)
- AAAI 2024 STAR: Boosting Low-Resource Event Extraction via Structure-to-Text Generation [Paper]
- EMNLP Findings 2023 Learning under Label Proportions for Text Classification [Paper]
- Preprint 2022 Global Responses to the COVID-19 Pandemic: Evidence Finding and Verification
- Speech Communication 2022 Seamless Equal Accuracy Ratio for Inclusive CTC Speech Recognition [Paper]
Invited Talks
-
ARLArena: A Unified Framework for Stable Agentic Reinforcement Learning March 2026NICE (Nexus for IntelligenCE)
Internship Experience
-
Preference Model Summer 2026Research Intern, Post-Training
-
Meta Platforms, Inc. Summer 2025Research Intern (AI)
-
Amazon Development Center U.S., Inc. Summer 2024Applied Scientist Intern
Education
-
University of California, Los Angeles Sept. 2022 – PresentPh.D. in Computer Science · Advisor: Wei Wang
-
University of Illinois Urbana-Champaign Aug. 2018 – May 2022B.S. in Computer Science, Grainger College of EngineeringMinor in Mathematics · James Scholar Honor · Dean's List
Teaching
-
CS 245: Big Data Analytics, UCLA Fall 2024Teaching Assistant
-
CS 446: Machine Learning, UIUC Spring 2022Course Assistant