Yafeng Sun

Shaobo Wang, Yujie Chen, Yafeng Sun et al.

Modern reasoning agents are increasingly evaluated on their ability to generate multiple valid solution paths, plans, or tool-use traces for a given input. Standard reward-maximizing RL tends to collapse onto the most easily reinforced high-reward mode, whereas distribution-matching RL aims to allocate probability mass across the entire reward-shaped solution set. Achieving this objective requires computing a prompt-dependent partition function over the trajectory space. Because existing distribution-matching methods learn this partition function online alongside the policy, calibration errors in the partition function directly distort policy updates and remain impossible to diagnose independently. We introduce DISA, short for Decoupled Importance-Sampled Anchoring, which moves this calibration problem outside the RL loop. DISA draws proposal trajectories offline, estimates the partition function via importance sampling, and freezes the resulting partition-function estimate before policy optimization begins. This decoupling preserves the distribution-matching objective while strictly separating partition-function estimation from policy learning in data, gradients, loss, and diagnostics. Empirically, on two open-weight backbones across six math and three code benchmarks, DISA matches or exceeds the online-coupled distribution-matching baseline FlowRL, outperforms rewardmaximization baselines GRPO and GSPO on math averages, and exceeds LoRASFT distillation by up to 13.8 Mean@8 points on the same offline trajectories. An LLM-as-judge evaluation further shows that DISA retains substantially more strategy-level diversity than reward-maximization baselines, and sensitivity studies on the proposal strength and inverse temperature follow the bias-variance pattern predicted by the analysis.

Runwei Guan, Ningwei Ouyang, Tianhao Xu et al.

Automated waterway environment perception is crucial for enabling unmanned surface vessels (USVs) to understand their surroundings and make informed decisions. Most existing waterway perception models primarily focus on instance-level object perception paradigms (e.g., detection, segmentation). However, due to the complexity of waterway environments, current perception datasets and models fail to achieve global semantic understanding of waterways, limiting large-scale monitoring and structured log generation. With the advancement of vision-language models (VLMs), we leverage image captioning to introduce WaterCaption, the first captioning dataset specifically designed for waterway environments. WaterCaption focuses on fine-grained, multi-region long-text descriptions, providing a new research direction for visual geo-understanding and spatial scene cognition. Exactly, it includes 20.2k image-text pair data with 1.8 million vocabulary size. Additionally, we propose Da Yu, an edge-deployable multi-modal large language model for USVs, where we propose a novel vision-to-language projector called Nano Transformer Adaptor (NTA). NTA effectively balances computational efficiency with the capacity for both global and fine-grained local modeling of visual features, thereby significantly enhancing the model's ability to generate long-form textual outputs. Da Yu achieves an optimal balance between performance and efficiency, surpassing state-of-the-art models on WaterCaption and several other captioning benchmarks.

Shaohan Sha, YaFeng Sun

Federated learning is a decentralized learning paradigm wherein a central server trains a global model iteratively by utilizing clients who possess a certain amount of private datasets. The challenge lies in the fact that the client side private data may not be identically and independently distributed, significantly impacting the accuracy of the global model. Existing methods commonly address the Non-IID challenge by focusing on optimization, client selection and data complement. However, most approaches tend to overlook the perspective of the private data itself due to privacy constraints.Intuitively, statistical distinctions among private data on the client side can help mitigate the Non-IID degree. Besides, the recent advancements in dataset condensation technology have inspired us to investigate its potential applicability in addressing Non-IID issues while maintaining privacy. Motivated by this, we propose DCFL which divides clients into groups by using the Centered Kernel Alignment (CKA) method, then uses dataset condensation methods with non-IID awareness to complete clients. The private data from clients within the same group is complementary and their condensed data is accessible to all clients in the group. Additionally, CKA-guided client selection strategy, filtering mechanisms, and data enhancement techniques are incorporated to efficiently and precisely utilize the condensed data, enhance model performance, and minimize communication time. Experimental results demonstrate that DCFL achieves competitive performance on popular federated learning benchmarks including MNIST, FashionMNIST, SVHN, and CIFAR-10 with existing FL protocol.