Hanqing Wang

Hanqing Wang

Final-Year Ph.D. Candidate at Shanghai University of Finance and Economics (SUFE)

I am a final-year Ph.D. candidate at SUFE advised by Assoc. Prof. Yun Chen. Previously, I was an algorithm intern at JD Technology working on VLA and embodied AI, and a research intern at THUNLP & Qiyuan Lab mentored by Shuo Wang. My work has appeared at NeurIPS, ACL, EMNLP, and NAACL, and I received the SUFE Academic Star (5 recipients annually) and the National Ph.D. Scholarship.

On the 2026–2027 job market: open to Postdoc and Research Scientist positions · Expected graduation: June 2027

My research focuses on Vision-Language-Action (VLA) models, efficient LLM adaptation, and LLM agents, driven by a unifying question: how can foundation models efficiently acquire, compose, and transfer capabilities across tasks, modalities, and physical environments? Along this thread, I have developed parameter-efficient adaptation and compression methods (StyleBART, LoRA-Flow, Delta-CoMe, MiLoRA), training-free cross-modal reasoning transfer from text LLMs to MLLMs (XPD), and physical dynamics representations for low-latency robotic manipulation (PDRVLA). In parallel, I study self-evolving and stateful LLM agents — from automated latent-vulnerability discovery and cross-language code transfer (SAGE, Pi-SQL) to stateful execution harnesses for long-horizon tasks (Perdura).

  • Oct 2026Released PDRVLA on arXiv — selected as Hugging Face Daily Paper #2 and Monthly Paper #4!
  • Oct 2026Served as Area Chair for ACL Rolling Review (ARR, Oct 2026 cycle).
  • Jul 2026Joined JD Technology as an Algorithm Intern working on VLA & Embodied AI.
  • Apr 2026SAGE accepted to Findings of ACL 2026!
  • Dec 2025Selected as SUFE Academic Star (5 recipients university-wide annually).
  • Nov 2025Pi-SQL accepted to Findings of EMNLP 2025!
  • Sep 2025Awarded National Scholarship for Doctoral Students.
  • Mar 2025Two papers (MiLoRA at Main, MALoRA at Findings) accepted to NAACL 2025!
  • Jan 2025Completed research internship at THUNLP & Qiyuan Lab.
  • Sep 2024Delta-CoMe accepted to NeurIPS 2024!
  • Sep 2024Awarded First-Class Academic Scholarship at SUFE.
  • May 2024LoRA-Flow accepted to ACL 2024 Main!
  • Oct 2023StyleBART accepted to Findings of EMNLP 2023!
  • Sep 2023Started research internship at THUNLP & Qiyuan Lab.
  • Jun 2022Received Bachelor's degree in Computer Science (Honors & Outstanding Graduate) from SUFE.
PDRVLA architecture and benchmark performance arXiv 2026
Hanqing Wang*, Zhaotian Cai*, Ruichuan An*, Bohan Zeng*, Tao Zhang*, Yilin Chen*, Xu Jiang, Kaiquan Chen, Chongyang Wu, Haoran Li, Fan Yang, Mingan Lin, Wentao Zhang, Zenan Zhou, Youzheng Wu, Xiaodong He
Paper · HF Paper · HF Daily Paper #2 · Monthly Paper #4
TL;DR: Extracts compact physical tokens from intermediate VLM features via direct future RGB/depth/optical-flow prediction and frozen video world-model supervision — conditioning VLA action generation without online video latency (99.3% LIBERO, 89.9% LIBERO-Plus, 93.6% RoboTwin 2.0, 81.5% SimplerEnv).
SAGE method overview ACL 2026
Hanqing Wang*, Yongdong Chi*, Jian Yang, Lei Yang, Jiehui Zhao, Yun Chen, Guanhua Chen
Paper · Code · ACL 2026 Findings · CCF-A
TL;DR: SAGE (Systematic Automated Guided Exploration) automates discovery of latent failure patterns in LLM Text-to-SQL — generating vulnerability hypotheses with a continuously evolving Vulnerability Codex of perturbations.
Pi-SQL pivot programming language pipeline EMNLP 2025
Yongdong Chi*, Hanqing Wang*, Yun Chen, Yan Yang, Jian Yang, Zonghan Yang, Xiao Yan, Guanhua Chen
Paper · Code · EMNLP 2025 Findings · CCF-B
TL;DR: Uses Python as a pivot intermediate representation to give Text-to-SQL fine-grained structural guidance — leveraging LLMs' stronger Python understanding to bootstrap better SQL generation.
MiLoRA minor singular component initialization NAACL 2025
Hanqing Wang, Yixia Li, Shuo Wang, Guanhua Chen, Yun Chen
Paper · Code · NAACL 2025 Main · CCF-B · 150+ citations
TL;DR: Initializes LoRA adapters using the minor singular components of pre-trained weights — reducing interference with the base model's existing knowledge while still enabling effective task adaptation.
Delta-CoMe mixed-precision delta compression NeurIPS 2024
Bowen Ping*, Shuo Wang*, Hanqing Wang, Xu Han, Yuzhuang Xu, Yukun Yan, Yun Chen, Baobao Chang, Zhiyuan Liu, Maosong Sun
Paper · Code · NeurIPS 2024 · CCF-A
TL;DR: Compresses the delta (fine-tuned minus base weights) of an aligned LLM with mixed-precision quantization — recovering near-full performance with a fraction of storage and zero retraining.
LoRA-Flow dynamic LoRA fusion ACL 2024
Hanqing Wang*, Bowen Ping*, Shuo Wang, Xu Han, Yun Chen, Zhiyuan Liu, Maosong Sun
Paper · Code · ACL 2024 Main · CCF-A
TL;DR: Replaces fixed LoRA fusion weights with a per-token dynamic gate — letting multiple task-specific LoRA adapters be combined adaptively during generation rather than averaged with a static recipe.
StyleBART style adapter architecture EMNLP 2023
Hanqing Wang*, Yajing Luo*, Boya Xiong, Guanhua Chen, Yun Chen
Paper · Code · EMNLP 2023 Findings · CCF-B
TL;DR: Adds lightweight style adapters on top of frozen BART to generate stylistic news headlines without paired data — decoupling content faithfulness from style transfer.