arXiv subjects

Yajun Wei

Publications and source records attributed to Yajun Wei.

5 recordsLinked to original sources

arXiv2026

DM0: An Embodied-Native Vision-Language-Action Model towards Physical AI

Moving beyond the traditional paradigm of adapting internet-pretrained models to physical tasks, we present DM0, an Embodied-Native Vision-Language-Action (VLA) framework designed for Physical AI. Unlike approaches that treat physical grounding as a fine-tuning afterthought, DM0 unifies embodied manipulation and navigation by learning from heterogeneous data sources from the onset. Our methodology follows a comprehensive three-stage pipeline: Pretraining, Mid-Training, and Post-Training. First, we conduct large-scale unified pretraining on the Vision-Language Model (VLM) using diverse corpora--seamlessly integrating web text, autonomous driving scenarios, and embodied interaction logs-to jointly acquire semantic knowledge and physical priors. Subsequently, we build a flow-matching action expert atop the VLM. To reconcile high-level reasoning with low-level control, DM0 employs a hybrid training strategy: for embodied data, gradients from the action expert are not backpropagated to the VLM to preserve generalized representations, while the VLM remains trainable on non-embodied data. Furthermore, we introduce an Embodied Spatial Scaffolding strategy to construct spatial Chain-of-Thought (CoT) reasoning, effectively constraining the action solution space. Experiments on the RoboChallenge benchmark demonstrate that DM0 achieves state-of-the-art performance in both Specialist and Generalist settings on Table30.

cs.RO↗

arXiv2025

RoboChallenge: Large-scale Real-robot Evaluation of Embodied Policies

Testing on real machines is indispensable for robotic control algorithms. In the context of learning-based algorithms, especially VLA models, demand for large-scale evaluation, i.e. testing a large number of models on a large number of tasks, is becoming increasingly urgent. However, doing this right is highly non-trivial, especially when scalability and reproducibility is taken into account. In this report, we describe our methodology for constructing RoboChallenge, an online evaluation system to test robotic control algorithms, and our survey of recent state-of-the-art VLA models using our initial benchmark Table30.

cs.RO↗

arXiv2025

Can Contextualized Physics Problems Enhance Student Motivation? Uncovering a Student-Teacher Perception Gap through a Large-Scale Survey

Embedding physics problems unreal-world settings, here termed contextualized physics problems (CPP), is widely believed to foster students' interest, motivation, and learning. However, firm evidence for this claim remains scarce. To explore this issue, we surveyed 868 secondary students and 154 teachers to examine their attitudes toward CPP and investigate whether students and teachers perceive these problems as promoting student interest and motivation in learning physics. The findings reveal a divergence between teacher and student perspectives. While most teachers view CPP as essential for enhancing interest and motivation, student responses tell a different story. Contextualized problems appear to boost interest and motivation only among 8th graders who are newly introduced to the subject. From 9th to 11th grade, students expressed a clear preference for de-contextualized physics problems (DPP) and generally disagreed that CPP increased their interest in physics. Gender differences were also observed among younger students, with boys showing a moderately stronger preference for CPP than girls. These results provide valuable insights for educators in designing course materials and creating effective test and exercise questions. The discrepancy between teacher and student perceptions, as highlighted by the surveys and interviews, underscores the need to address this gap through targeted teacher training and professional development.

physics.ed-ph↗

arXiv2023

An entertaining resonance experiment with just two spring scales

Resonance is a topic included in most introductory physics courses. Any mechanical system experiences resonance if it is driven by a periodic force with a frequency that matches its natural frequency. There are plenty of simple demonstrations of the resonance phenomena of mechanical systems which can be set up using readily available items . This work presents a very simple approach to demonstrate the phenomena using just two spring scales. The experiment presented here performs a "frequency sweep" and is also very entertaining to watch.

physics.ed-ph↗

arXiv2023

Novel Approaches to Solve Simple Harmonic Motion

This paper presents two novel approaches to solve the classic simple harmonic motion. In one approach, the distance between the equilibrium position and the maximal displacement is divided into N equal segments. In each segment, the mass moves with constant acceleration under the average of two forces at the ends of the segment. Summing up the time covering each segment and taking the large-N limit reproduce one quarter of the period for simple harmonic motion. In the other approach, the time moving from the maximal displacement to the equilibrium position is divided into N equal intervals. A recurrence relation for the displacement is obtained. The large-N limit of its solution results in the same solution as that obtained from solving differential equation.

physics.class-ph↗