CV

Maowei Jiang / 蒋茂苇

Email: jmw24@mails.tsinghua.edu.cn
GitHub: Zero-coder
Google Scholar: Profile

Research Interests

Large Language Models, Multimodal Learning, Vision-Language-Action Models, Robot Learning, Embodied World Models, Reinforcement Learning, Policy Optimization, Long-Horizon Decision Making, LLM/VLM Agents.

Education

• Tsinghua University, School of Future Human Habitats, 2024.09 - Present
  Research focus: Large models, agents, embodied intelligence.
• University of Chinese Academy of Sciences / Shenyang Institute of Automation, Chinese Academy of Sciences, 2021.08 - 2024.06
  Research focus: Deep learning and intelligent perception.
• Mianyang Teachers’ College, School of Mechanical and Electrical Engineering, 2017.07 - 2021.06
  Research focus: Electrical engineering and embedded automation.

Selected Publications and Submissions

1. TAPO: Dynamic Teacher and Perturbed Answer Injection for Policy Optimization
   AAAI 2026 Oral. First author. LLM policy optimization with dynamic teacher signals and perturbed answer injection.

2. FutureVLA: Acting on Predicted Futures with Vision-Language-Action Models
   NeurIPS 2026 under review. First author. Future-conditioned VLA framework connecting visual world-model prediction and robot action generation.

3. ReCon: Reference-Conditioned Online Refinement for Vision-Language-Action Policies
   NeurIPS 2026 under review. First author. Online residual correction for frozen VLA policies on real-robot contact-rich tasks.

4. Prompt2Act: Mapping Prompts into Sequence of Robotic Actions with Large Foundation Models
   Information Fusion, IF 15.5, Q1 Top, CCF-B. First author. Natural-language prompts to executable robot action sequences.

5. FDVLA: A Flow-Diffusion Vision-Language-Action Framework with Dual Reasoning Modulation
   Information Fusion under review. First author. Flow-diffusion VLA modeling and reasoning modulation for robot manipulation.

6. RL2VLA: Reinforcement Learning Fine-tuning for Vision-Language-Action Models
   ACM MM 2026 under review. First author. Reinforcement learning fine-tuning for VLA models.

7. DAAC: Discrepancy-Aware Adaptive Contrastive Learning
   NeurIPS 2025. Co-first / project lead. Adaptive contrastive learning under distribution discrepancy.

8. CARD / MRED-14 / GreenPlanner
   NeurIPS 2024 Workshop on Open-World Agents; ACM MM; CVPR. Contributions to cross-modal agents, generative residential design, and constraint-aware layout generation.

Research and Engineering Experience

• Embodied AI and VLA: Future-conditioned VLA, online policy refinement, prompt-to-action generation, robot action representation, contact-rich real-robot tasks, long-horizon decision making.
• LLM / Agent / RL: SFT, LoRA, GRPO, PPO, SAC, reward design, reward hacking analysis, policy optimization, agent evaluation, credit assignment.
• Multimodal Learning: LLaMA / LLaVA-style fine-tuning, modality alignment, multimodal representation learning, visual grounding, future-state prediction.
• Engineering: Python, PyTorch, Transformers, DeepSpeed, TRL / VERL, Docker, Git, Linux, multi-GPU training, reproducible experiment management.

Open Source

• Awesome-LLM-Robotics: contributor to a 4.4k+ star resource list for LLM/MLLM + Robotics/RL.
• Second-Me: contributor to a 15.5k+ star personalized AI self system.
• Prompt2Act, RL2VLA, FDVLA: research repositories for VLA and robot action generation.
• Zero-coder: 98 public repositories covering LLMs, VLA, multimodal learning, and computer vision.

Honors and Competitions

• Kaggle CMI Child Mind Institute: Silver Medal, global rank 75 / 1878.
• Huawei Ascend AI Innovation Competition: Excellent Solution Award.
• BMW Hackathon: Finalist / second place.
• Alibaba Tianchi Few-shot Trademark Detection: Global rank 239 / 2135.
• Asia-Pacific Ophthalmology Big Data Competition: Global rank 142 / 10006.