Point-based geometric representations such as point clouds and Gaussian Splatting are fundamental for 3D understanding. However, the inherent irregularity and high-dimensional nature of point structures present significant challenges for direct 3D learning approaches, which often struggle with scalability and achieve suboptimal performance due to sparse data distributions. In contrast, 2D learning paradigms benefit from well-established architectures with superior optimization stability and efficiency. To bridge this gap, we propose Maniflat3D, a unified framework that systematically transforms volumetric point-based geometries into structured 2D representations through a two-stage process: a multilayer Ball-Pivoting reconstruction with adaptive density control, followed by Scalable Locally Injective Mapping (SLIM) to produce distortion-minimized, bijective UV parameterizations. Our approach explicitly encodes both geometric and attribute information into the flattened domain, enabling conventional 2D neural networks to effectively learn from complex 3D structures such as Gaussian Splatting. Experiments on the ShapeSplat dataset demonstrate that Maniflat3D achieves comparable performance while reducing parameter count by 90% compared to native 3D baselines, and simultaneously attains 21× compression ratio through neural encoding. These results establish a new paradigm for efficient geometric understanding, demonstrating successful transfer of planar learning advantages to challenging 3D manifold problems through dimensional reduction.

We introduce EURUS, a suite of large language models (LLMs) optimized for reasoning. Finetuned from Mistral-7B, Llama-3-8B, and Mixtral-8x22B, EURUS models achieve state-of-the-art results among open-source models on a diverse set of benchmarks covering mathematics, code generation, and logical reasoning problems. Notably, EURUX-8X22B outperforms GPT-3.5 Turbo in reasoning through a comprehensive benchmarking across 12 test sets covering five tasks. The strong performance of EURUS can be primarily attributed to ULTRAINTERACT, our newly-curated large-scale, high-quality training data dataset specifically designed for complex reasoning tasks. ULTRAINTERACT can be used in both supervised fine-tuning, preference learning, and reward modeling. It pairs each instruction with a preference tree consisting of (1) reasoning chains with diverse planning strategies in a unified format, (2) multi-turn interaction trajectories with the environment and the critique, and (3) pairwise positive and negative responses to facilitate preference learning. ULTRAINTERACT allows us to conduct an in-depth exploration of preference learning for reasoning tasks. Our investigation reveals that some well-established preference learning algorithms may be less suitable for reasoning tasks compared to their effectiveness in general conversations. The hypothesis is that in reasoning tasks, the space of correct answers is much smaller than that of incorrect ones, so it is necessary to explicitly increase the reward of chosen data. Therefore, in addition to increasing the reward margin as many preference learning algorithms do, the absolute values of positive responses’ rewards should be positive and may serve as a proxy for performance. Inspired by this, we derive a novel reward modeling objective and empirically that it leads to a stable reward modeling curve and better performance. Together with ULTRAINTERACT, we obtain a strong reward model.

Zeyuan Liu, Ziyu Huan, Xiyao Wang, Jiafei Lyu, Jian Tao, Xiu Li, Furong Huang, Huazhe Xu. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Real-world datasets collected from sensors or human inputs are prone to noise and errors, posing significant challenges for applying offline reinforcement learning (RL). While existing methods have made progress in addressing corrupted actions and rewards, they remain insufficient for handling corruption in high-dimensional state spaces and for cases where multiple elements in the dataset are corrupted simultaneously. Diffusion models, known for their strong denoising capabilities, offer a promising direction for this problem—but their tendency to overfit noisy samples limits their direct applicability. To overcome this, we propose **A**mbient **D**iffusion-**G**uided Dataset Recovery (**ADG**), a novel approach that pioneers the use of diffusion models to tackle data corruption in offline RL. First, we introduce Ambient Denoising Diffusion Probabilistic Models (DDPM) from approximated distributions, which enable learning on partially corrupted datasets with theoretical guarantees. Second, we use the noise-prediction property of Ambient DDPM to distinguish between clean and corrupted data, and then use the clean subset to train a standard DDPM. Third, we employ the trained standard DDPM to refine the previously identified corrupted data, enhancing data quality for subsequent offline RL training. A notable strength of ADG is its versatility—it can be seamlessly integrated with any offline RL algorithm. Experiments on a range of benchmarks, including MuJoCo, Kitchen, and Adroit, demonstrate that ADG effectively mitigates the impact of corrupted data and improves the robustness of offline RL under various noise settings, achieving state-of-the-art results.

Ensuring fairness in machine learning models is a critical challenge. Existing debiasing methods often compromise performance, rely on static correction strategies, and struggle with data sparsity, particularly within minority groups. Furthermore, their utilization of sensitive attributes is often suboptimal, either depending excessively on complete attribute labeling or disregarding these attributes entirely. To overcome these limitations, we propose FairNet, a novel framework for dynamic, instance-level fairness correction. FairNet integrates a bias detector with conditional low-rank adaptation (LoRA), which enables selective activation of the fairness correction mechanism exclusively for instances identified as biased, and thereby preserve performance on unbiased instances. A key contribution is a new contrastive loss function for training the LoRA module, specifically designed to minimize intra-class representation disparities across different sensitive groups and effectively address underfitting in minority groups. The FairNet framework can flexibly handle scenarios with complete, partial, or entirely absent sensitive attribute labels. Theoretical analysis confirms that, under moderate TPR/FPR for the bias detector, FairNet can enhance the performance of the worst group without diminishing overall model performance, and potentially yield slight performance improvements. Comprehensive empirical evaluations across diverse vision and language benchmarks validate the effectiveness of FairNet. Code is available at \url{https://github.com/SongqiZhou/FairNet}.

The partial observability and stochasticity in multi-agent settings can be mitigated by accessing more information about others via communication. However, the coordination problem still exists since agents cannot communicate actual actions with each other at the same time due to the circular dependencies. In this paper, we propose a novel multi-level communication scheme, Sequential Communication (SeqComm). SeqComm treats agents asynchronously (the upper-level agents make decisions before the lower-level ones) and has two communication phases. In the negotiation phase, agents determine the priority of decision-making by communicating hidden states of observations and comparing the value of intention, which is obtained by modeling the environment dynamics. In the launching phase, the upper-level agents take the lead in making decisions and then communicate their actions with the lower-level agents. Theoretically, we prove the policies learned by SeqComm are guaranteed to improve monotonically and converge. Empirically, we show that SeqComm outperforms existing methods in a variety of cooperative multi-agent tasks.

Federated learning is exposed to model poisoning attacks as compromised clients may submit malicious model updates to pollute the global model. To defend against such attacks, robust aggregation rules are designed for the centralized server to winnow out outlier updates, and to significantly reduce the effectiveness of existing poisoning attacks. In this paper, we develop an advanced model poisoning attack against defensive aggregation rules. In particular, we exploit the catastrophic forgetting phenomenon during the process of continual learning to destroy the memory of the global model. Our proposed framework, called Oblivion, features two special components. The first component prioritizes the weights that have the most influence on the model accuracy for poisoning, which induces a more significant degradation on the global model than equally perturbing all weights. The second component smooths malicious model updates based on the number of selected compromised clients in the current round, adjusting the degree of poisoning to suit the dynamics of each training round. We implement a fully-functional prototype of Oblivion in PLATO, a real-world scalable federated learning framework. Our extensive experiments over three datasets demonstrate that Oblivion can boost the attack performance of model poisoning attacks against unknown defensive aggregation rules.

A novel single-winding bearingless switched reluctance motor with wider rotor teeth (BSRMWRW) is proposed in this article, which can realize the decoupled control of torque and levitation force and thus simplify the motor's control in the starter/generator system. First, the working principles and decoupling characteristic of BSRMWR are explained; Second, the mathematical model of torque and levitation force is derived based on the Maxwell stress tensor method. The model's accuracy is further verified by the finite element analysis (FEA); Based on this model, the square-current control method is proposed with average torque oriented. In addition, the proposed suspension control strategy not only reduces the number of system variables but also eliminate the coupling of levitation forces in X and Y directions. Finally, the simulation of single-winding BSRMWR is carried out in Matlab/Simulink.

Non-orthogonal multiple access (NOMA) is an emerging research topic in the future fifth generation wireless communication networks, which is expected to support massive connectivity for massive machine-type communications (mMTC). Due to the sporadic communication nature of mMTC, the grant-free transmission methodology is highly expected in uplink NOMA systems, to drastically reduce the transmission latency and signaling overhead. Exploiting the inherent sparsity nature of user activity, compressive sensing (CS) techniques have been applied for efficient multi-user detection in the uplink grant-free NOMA. In this paper, we propose a prior-information-aided adaptive subspace pursuit (PIA-ASP) algorithm to improve the multi-user detection performance. In this algorithm, a parameter evaluating the quality of the prior-information support set is introduced, in order to exploit the intrinsically temporal correlation of active user support sets in several continuous time slots adaptively. Then, to mitigate the incorrect estimation effect of the prior support quality information, a robust PIA-ASP algorithm is further proposed, which adaptively exploits the prior support based on the corresponding support quality information in a conservative way. It is noted that both of the two proposed algorithms do not require the knowledge of the user sparsity level, while most of the state-of-the-art CS-based multi-user detection algorithms usually need. Moreover, for the two proposed algorithms, the upper bound of the signal detection error and the computational complexity is derived. Simulation results demonstrate that the two proposed algorithms are capable of achieving much better performance than that of the existing CS-based multi-user detection algorithms with a similar computational complexity.