01 — About

About

Embodied AI researcher working on vision-language-action models, reinforcement learning and post-training for real robots.

I am a Master's student in Electrical Engineering at the National University of Singapore (NUS) and a core member of the NUS CORE Lab (COntrol, Robotics & Artificial IntElligence). I work on Embodied AI — how robots perceive, understand language and act in the physical world. What I care about most is the whole loop: not just the algorithm, but the data, the infrastructure and the deployment that turn it into behaviour on a real robot.

In summer 2026 (Jun 1 – Aug 31) I was a VLA Algorithm Engineer (Intern) on Honor's humanoid-robot team in Shanghai, where I took part in building the direction from the ground up and owned several of its core pieces. I led the teleoperation and data pipeline — smooth human takeover during autonomous rollouts, on-robot recording with automatic upload — and closed a distributed real-robot SFT / RL loop across the robot's onboard boards, an x86 actor and a learner server. With that infrastructure in place I designed the experiments myself: reproducing RECAP-style offline RL and combining it with noise-space fast adaptation, which substantially raised success rates on tasks the offline data never covered.

My research has produced two IROS 2026 papers within a year. As first author of LangGap I showed that state-of-the-art VLA models score highly while largely ignoring language, and proposed a semantic-perturbation benchmark plus same-scene multi-task training to close the gap; in AION I designed and trained the vision-only exploration policy for zero-shot aerial object navigation. Before NUS I studied Electronic and Communication Engineering at Nanjing University of Aeronautics and Astronautics (NUAA).

02 — NewsAll news

Aug 31, 2026
Completed my VLA algorithm internship at Honor (humanoid robot).
Jun 17, 2026
🎉 Two papers accepted to IROS 2026: LangGap (first author) and AION (co-author).

03 — Selected publicationsAll publications

IROS 2026 11 citations

LangGap: Diagnosing and Closing the Language Gap in Vision-Language-Action Models

Yuchen Hou and Lin Zhao

In 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2026

Abs arXiv HTML Poster

Vision-Language-Action (VLA) models achieve over 95% success on standard benchmarks. However, through systematic experiments, we find that current state-of-the-art VLA models largely ignore language instructions. Prior work lacks: (1) systematic semantic perturbation diagnostics, (2) a benchmark that forces language understanding by design, and (3) linguistically diverse training data. This paper constructs the LangGap benchmark, based on a four-dimensional semantic perturbation method – varying instruction semantics while keeping the tabletop layout fixed – revealing language understanding deficits in \pi0.5. Existing benchmarks like LIBERO assign only one task per layout, underutilizing available objects and target locations; LangGap fully diversifies pick-and-place tasks under identical layouts, forcing models to truly understand language. Experiments show that targeted data augmentation can partially close the language gap – success rate improves from 0% to 90% with single-task training, and 0% to 28% with multi-task training. However, as semantic diversity of extended tasks increases, model learning capacity proves severely insufficient; even trained tasks perform poorly. This reveals a fundamental challenge for VLA models in understanding diverse language instructions – precisely the long-term value of LangGap.
IROS 2026 1 citations

AION: Aerial Indoor Object-Goal Navigation Using Dual-Policy Reinforcement Learning

Zichen Yan, Yuchen Hou, Shenao Wang, and 3 more authors

In 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2026

Abs arXiv HTML Video

Object-Goal Navigation (ObjectNav) requires an agent to autonomously explore an unknown environment and navigate toward target objects specified by a semantic label. While prior work has primarily studied zero-shot ObjectNav under 2D locomotion, extending it to aerial platforms with 3D locomotion capability remains underexplored. Aerial robots offer superior maneuverability and search efficiency, but they also introduce new challenges in spatial perception, dynamic control, and safety assurance. In this paper, we propose AION for vision-based aerial ObjectNav without relying on external localization or global maps. AION is an end-to-end dual-policy reinforcement learning (RL) framework that decouples exploration and goal-reaching behaviors into two specialized policies. We evaluate AION on the AI2-THOR benchmark and further assess its real-time performance in IsaacSim using high-fidelity drone models. Experimental results show that AION achieves superior performance across comprehensive evaluation metrics in exploration, navigation efficiency, and safety.

04 — Selected workAll projects

Jun 1 – Aug 31, 2026 Honor Device Co., Ltd. · Summer internship Honor Humanoid: Real-Robot RL for VLA VLA Algorithm Engineer Intern at Honor. Built a human-in-the-loop real-robot SFT / RL pipeline for Honor's self-developed humanoid robot — smooth teleoperator takeover, an intervention-driven data pipeline, and a distributed Actor–Learner–Robot learning loop.

IROS 2026 11citations Oct 2025 – Jun 2026 First author LangGap: VLA Language Understanding Benchmark Accepted at IROS 2026. Designed a systematic semantic perturbation evaluation framework revealing that state-of-the-art VLA models ignore language instructions despite high benchmark scores. Proposed multi-task same-scene training approach and constructed augmented dataset for fine-tuning.

IROS 2026 1citations Jun 2025 – Jun 2026 Co-author AION: Aerial Indoor Object-Goal Navigation Accepted at IROS 2026. End-to-end dual-policy RL framework for vision-based aerial ObjectNav without external localization or global maps. Evaluated on AI2-THOR and IsaacSim.

Jun 2025 – Feb 2026 Habitat · Aerial VLN dataset pipeline Vision-Language Navigation on Autonomous Drone Built a robust pipeline to generate various 3D paths in the Habitat simulator. Overcame challenges of the simulator initially designed only for ground robots by designing a robust 3D navigation algorithm and obstacle detection method. Trained a strong and general policy for drone navigation.

05 — Contact

I'm looking for research positions in industry labs working on embodied AI and robot learning, and am equally open to PhD opportunities and academic collaborations. Reach me at yuchenhou287@outlook.com.