Yifan Shen

Yifan Shen


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 April 2025 in Proceedings of the AAAI Conference on Artificial Intelligence
Verified
Journal Article
Open access
PDF file
EXEric Xue
Eric Xue
YLYijiang Li
Yijiang Li
HLHaoyang Liu
Haoyang Liu
PWPeiran Wang
Peiran Wang
YSYifan Shen
Yifan Shen
HWHaohan Wang
Haohan Wang
Dataset distillation (DD) allows datasets to be distilled to fractions of their original size while preserving the rich distributional information so that models trained on the distilled datasets can achieve a comparable accuracy while saving significant computational loads. Recent research in this area has been focusing on improving the accuracy of models trained on distilled datasets. In this paper, we aim to explore a new perspective of DD. We study how to embed adversarial robustness in distilled datasets, so that models trained on these datasets maintain the high accuracy and meanwhile acquire better adversarial robustness. We propose a new method that achieves this goal by incorporating curvature regularization into the distillation process with much less computational overhead than standard adversarial training. Extensive empirical experiments suggest that our method not only outperforms standard adversarial training on both accuracy and robustness with less computation overhead but is also capable of generating robust distilled datasets that can withstand various adversarial attacks.


2 June 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
HZHaichao Zhang
Haichao Zhang
MCMingfei Chen
Mingfei Chen
SHShwai He
Shwai He
ZXZhengtong Xu
Zhengtong Xu
YSYifan Shen
Yifan Shen
YFYun Fu
Yun Fu
Physical AI aims to extend artificial intelligence from digital reasoning to perception, prediction, simulation, planning, and action in the physical world. While recent progress has advanced through vision-language models, vision-language-action models, world models, policy learning, and embodied agents, existing discussions are often organized from robotics-centric, vision-centric, or cyber-physical perspectives. This survey instead studies Physical AI through the lens of LLM-based world knowledge. We argue that LLMs encode implicit semantic, commonsense, procedural, and causal priors through large-scale pretraining, making them useful high-level sources for physical reasoning, multimodal grounding, action grounding, and embodied decision making. However, language-mediated knowledge is sparse and lossy for dense physical states, such as geometry, motion, contact, force, high-frequency dynamics, and long-horizon temporal evolution. VLMs and MLLMs ground LLM-derived priors into perception, but often expose physical understanding through language outputs. VLAs connect perception and language to executable actions, yet typically lack predictive models of how the world evolves under actions. This motivates a roadmap from LLM-based world knowledge to multimodal grounding, action grounding, world modeling, policy learning, and embodied deployment. We review how these components provide perceptual, actionable, predictive, and simulative substrates for Physical AI, and discuss open challenges in grounding, world modeling, closed-loop evaluation, safety, and generalization. A curated repository of related papers and resources is available at https://github.com/Hai-chao-Zhang/Awesome-Physical-AI.

19 October 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
YSYifan Shen
Yifan Shen
ZZZhuoqing Zhong
Zhuoqing Zhong
XLXinzhuo Li
Xinzhuo Li
JLJiateng Liu
Jiateng Liu
BLBingxuan Li
Bingxuan Li
ILIsmini Lourentzou
Ismini Lourentzou
The rapid progress of large language models has established strong foundations for reasoning, knowledge, and generation, while the integration of vision, audio, video, and other modalities has extended these capabilities toward increasingly unified intelligence across modalities, which we refer to as Omni Intelligence. Yet, real-world assistants must do more than process completed multimodal prompts. They must perceive evolving streams, respond while new evidence is arriving, handle hesitation, overlap, interruption, and changing intent, and coordinate reasoning or tool use that may proceed asynchronously. We refer to this broader capability as \( \textbf{Interactive Omni Intelligence} \): the ability to maintain and exercise omni intelligence under the temporal and adaptive demands of real-time interaction. The question guiding this paper is: How can foundation-model capabilities be carried into real-time interaction while preserving both intelligence and interactivity?We develop a conceptual framework connecting two development paths. One adapts pretrained language or omni models through streaming interfaces, modality modules, post-training, and orchestration. The other develops model-native systems that learn temporal structure, continuous multimodal processing, generation, and interaction control. Across both paths, we compare representations, architectures, fusion, data, training objectives, serving, and evaluation. Their comparison reveals an Intelligence--Interactivity Trade-off between preserving reasoning and knowledge, and meeting the timing, coordination, and resource demands of continuous operation. We examine Multi-Rate Interactive Intelligence as a cross-cutting direction, including foreground--background designs that coordinate fast interaction with slower reasoning, retrieval, memory, and tools. Finally, we identify open problems in evolving state, persistent and proactive interaction, joint evaluation, and safe open-world operation. The resulting framework connects capability preservation, interaction control, and multi-rate coordination.