Yueheng Li

Yueheng Li, Guangming Xie, Zongqing Lu

Due to practical constraints such as partial observability and limited communication, Centralized Training with Decentralized Execution (CTDE) has become the dominant paradigm in cooperative Multi-Agent Reinforcement Learning (MARL). However, existing CTDE methods often underutilize centralized training or lack theoretical guarantees. We propose Multi-Agent Guided Policy Optimization (MAGPO), a novel framework that better leverages centralized training by integrating centralized guidance with decentralized execution. MAGPO uses an auto-regressive joint policy for scalable, coordinated exploration and explicitly aligns it with decentralized policies to ensure deployability under partial observability. We provide theoretical guarantees of monotonic policy improvement and empirically evaluate MAGPO on 43 tasks across 6 diverse environments. Results show that MAGPO consistently outperforms strong CTDE baselines and matches or surpasses fully centralized approaches, offering a principled and practical solution for decentralized multi-agent learning. Our code and experimental data can be found in https://github.com/liyheng/MAGPO.

Yueheng Li, Guangming Xie, Zongqing Lu

Reinforcement Learning (RL) in partially observable environments poses significant challenges due to the complexity of learning under uncertainty. While additional information, such as that available in simulations, can enhance training, effectively leveraging it remains an open problem. To address this, we introduce Guided Policy Optimization (GPO), a framework that co-trains a guider and a learner. The guider takes advantage of privileged information while ensuring alignment with the learner's policy that is primarily trained via imitation learning. We theoretically demonstrate that this learning scheme achieves optimality comparable to direct RL, thereby overcoming key limitations inherent in existing approaches. Empirical evaluations show strong performance of GPO across various tasks, including continuous control with partial observability and noise, and memory-based challenges, significantly outperforming existing methods.

Jing Liu, Haozheng Wang, Yueheng Li

Large language models struggle with representing and generating rare tokens despite their importance in specialized domains. In this study, we identify neuron structures with exceptionally strong influence on language model's prediction of rare tokens, termed as rare token neurons, and investigate the mechanism for their emergence and behavior. These neurons exhibit a characteristic three-phase organization (plateau, power-law, and rapid decay) that emerges dynamically during training, evolving from a homogeneous initial state to a functionally differentiated architecture. In the activation space, rare token neurons form a coordinated subnetwork that selectively co-activates while avoiding co-activation with other neurons. This functional specialization potentially correlates with the development of heavy-tailed weight distributions, suggesting a statistical mechanical basis for emergent specialization.

Jing Liu, Haozheng Wang, Yueheng Li

Large language models (LLMs) struggle with representing and generating rare tokens despite their importance in specialized domains. We investigate whether LLMs develop internal specialization mechanisms through discrete modular architectures or distributed parameter-level differentiation. Through systematic analysis of final-layer MLP neurons across multiple model families, we discover that rare-token processing emerges via \textit{distributed specialization}: functionally coordinated but spatially distributed subnetworks that exhibit three distinct organizational principles. First, we identify a reproducible three-regime influence hierarchy comprising highly influential plateau neurons(also termed as rare-token neurons), power-law decay neurons, and minimally contributing neurons, which is absent in common-token processing. Second, plateau neurons demonstrate coordinated activation patterns (reduced effective dimensionality) while remaining spatially distributed rather than forming discrete clusters. Third, these specialized mechanisms are universally accessible through standard attention pathways without requiring dedicated routing circuits. Training dynamics reveal that functional specialization emerges gradually through parameter differentiation, with specialized neurons developing increasingly heavy-tailed weight correlation spectra consistent with Heavy-Tailed Self-Regularization signatures. Our findings establish that LLMs process rare-tokens through distributed coordination within shared architectures rather than mixture-of-experts-style modularity. These results provide insights for interpretable model editing, computational efficiency optimization, and understanding emergent functional organization in transformer networks.

Chuang Zhu, Shengjie Liu, Zekuan Yu et al.

For invasive breast cancer, immunohistochemical (IHC) techniques are often used to detect the expression level of human epidermal growth factor receptor-2 (HER2) in breast tissue to formulate a precise treatment plan. From the perspective of saving manpower, material and time costs, directly generating IHC-stained images from Hematoxylin and Eosin (H&E) stained images is a valuable research direction. Therefore, we held the breast cancer immunohistochemical image generation challenge, aiming to explore novel ideas of deep learning technology in pathological image generation and promote research in this field. The challenge provided registered H&E and IHC-stained image pairs, and participants were required to use these images to train a model that can directly generate IHC-stained images from corresponding H&E-stained images. We selected and reviewed the five highest-ranking methods based on their PSNR and SSIM metrics, while also providing overviews of the corresponding pipelines and implementations. In this paper, we further analyze the current limitations in the field of breast cancer immunohistochemical image generation and forecast the future development of this field. We hope that the released dataset and the challenge will inspire more scholars to jointly study higher-quality IHC-stained image generation.

Tianhao Zhang, Yueheng Li, Shuai Li et al.

In this paper, the circle formation control problem is addressed for a group of cooperative underactuated fish-like robots involving unknown nonlinear dynamics and disturbances. Based on the reinforcement learning and cognitive consistency theory, we propose a decentralized controller without the knowledge of the dynamics of the fish-like robots. The proposed controller can be transferred from simulation to reality. It is only trained in our established simulation environment, and the trained controller can be deployed to real robots without any manual tuning. Simulation results confirm that the proposed model-free robust formation control method is scalable with respect to the group size of the robots and outperforms other representative RL algorithms. Several experiments in the real world verify the effectiveness of our RL-based approach for circle formation control.