arXiv subjects

Yue Wang

Publications and source records attributed to Yue Wang.

At least 343 records · Page 19Linked to original sources

arXiv2025

The MOTIF Hand: A Robotic Hand for Multimodal Observations with Thermal, Inertial, and Force Sensors

Advancing dexterous manipulation with multi-fingered robotic hands requires rich sensory capabilities, while existing designs lack onboard thermal and torque sensing. In this work, we propose the MOTIF hand, a novel multimodal and versatile robotic hand that extends the LEAP hand by integrating: (i) dense tactile information across the fingers, (ii) a depth sensor, (iii) a thermal camera, (iv), IMU sensors, and (v) a visual sensor. The MOTIF hand is designed to be relatively low-cost (under 4000 USD) and easily reproducible. We validate our hand design through experiments that leverage its multimodal sensing for two representative tasks. First, we integrate thermal sensing into 3D reconstruction to guide temperature-aware, safe grasping. Second, we show how our hand can distinguish objects with identical appearance but different masses - a capability beyond methods that use vision only.

cs.RO↗

arXiv2025

Nature Language Model: Deciphering the Language of Nature for Scientific Discovery

Foundation models have revolutionized natural language processing and artificial intelligence, significantly enhancing how machines comprehend and generate human languages. Inspired by the success of these foundation models, researchers have developed foundation models for individual scientific domains, including small molecules, materials, proteins, DNA, RNA and even cells. However, these models are typically trained in isolation, lacking the ability to integrate across different scientific domains. Recognizing that entities within these domains can all be represented as sequences, which together form the "language of nature", we introduce Nature Language Model (NatureLM), a sequence-based science foundation model designed for scientific discovery. Pre-trained with data from multiple scientific domains, NatureLM offers a unified, versatile model that enables various applications including: (i) generating and optimizing small molecules, proteins, RNA, and materials using text instructions; (ii) cross-domain generation/design, such as protein-to-molecule and protein-to-RNA generation; and (iii) top performance across different domains, matching or surpassing state-of-the-art specialist models. NatureLM offers a promising generalist approach for various scientific tasks, including drug discovery (hit generation/optimization, ADMET optimization, synthesis), novel material design, and the development of therapeutic proteins or nucleotides. We have developed NatureLM models in different sizes (1 billion, 8 billion, and 46.7 billion parameters) and observed a clear improvement in performance as the model size increases.

cs.AI↗

arXiv2025

Trust Region Preference Approximation: A simple and stable reinforcement learning algorithm for LLM reasoning

Recently, Large Language Models (LLMs) have rapidly evolved, approaching Artificial General Intelligence (AGI) while benefiting from large-scale reinforcement learning to enhance Human Alignment (HA) and Reasoning. Recent reward-based optimization algorithms, such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO) have achieved significant performance on reasoning tasks, whereas preference-based optimization algorithms such as Direct Preference Optimization (DPO) significantly improve the performance of LLMs on human alignment. However, despite the strong performance of reward-based optimization methods in alignment tasks , they remain vulnerable to reward hacking. Furthermore, preference-based algorithms (such as Online DPO) haven't yet matched the performance of reward-based optimization algorithms (like PPO) on reasoning tasks, making their exploration in this specific area still a worthwhile pursuit. Motivated by these challenges, we propose the Trust Region Preference Approximation (TRPA) algorithm, which integrates rule-based optimization with preference-based optimization for reasoning tasks. As a preference-based algorithm, TRPA naturally eliminates the reward hacking issue. TRPA constructs preference levels using predefined rules, forms corresponding preference pairs, and leverages a novel optimization algorithm for RL training with a theoretical monotonic improvement guarantee. Experimental results demonstrate that TRPA not only achieves competitive performance on reasoning tasks but also exhibits robust stability. The code of this paper are released and updating on https://github.com/XueruiSu/Trust-Region-Preference-Approximation.git.

cs.LG↗

arXiv2025

TrialBench: Multi-Modal Artificial Intelligence-Ready Clinical Trial Datasets

Clinical trials are pivotal for developing new medical treatments but typically carry risks such as patient mortality and enrollment failure that waste immense efforts spanning over a decade. Applying artificial intelligence (AI) to predict key events in clinical trials holds great potential for providing insights to guide trial designs. However, complex data collection and question definition requiring medical expertise have hindered the involvement of AI thus far. This paper tackles these challenges by presenting a comprehensive suite of 23 meticulously curated AI-ready datasets covering multi-modal input features and 8 crucial prediction challenges in clinical trial design, encompassing prediction of trial duration, patient dropout rate, serious adverse event, mortality rate, trial approval outcome, trial failure reason, drug dose finding, design of eligibility criteria. Furthermore, we provide basic validation methods for each task to ensure the datasets' usability and reliability. We anticipate that the availability of such open-access datasets will catalyze the development of advanced AI approaches for clinical trial design, ultimately advancing clinical trial research and accelerating medical solution development.

cs.LG↗

arXiv2025

SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending

Humanoid robots hold significant potential in accomplishing daily tasks across diverse environments thanks to their flexibility and human-like morphology. Recent works have made significant progress in humanoid whole-body control and loco-manipulation leveraging optimal control or reinforcement learning. However, these methods require tedious task-specific tuning for each task to achieve satisfactory behaviors, limiting their versatility and scalability to diverse tasks in daily scenarios. To that end, we introduce SkillBlender, a novel hierarchical reinforcement learning framework for versatile humanoid loco-manipulation. SkillBlender first pretrains goal-conditioned task-agnostic primitive skills, and then dynamically blends these skills to accomplish complex loco-manipulation tasks with minimal task-specific reward engineering. We also introduce SkillBench, a parallel, cross-embodiment, and diverse simulated benchmark containing three embodiments, four primitive skills, and eight challenging loco-manipulation tasks, accompanied by a set of scientific evaluation metrics balancing accuracy and feasibility. Extensive simulated experiments show that our method significantly outperforms all baselines, while naturally regularizing behaviors to avoid reward hacking, resulting in more accurate and feasible movements for diverse loco-manipulation tasks in our daily scenarios. Our code and benchmark will be open-sourced to the community to facilitate future research. Project page: https://usc-gvl.github.io/SkillBlender-web/.

cs.RO↗

arXiv2025

Learning Temporally Consistent Video Depth from Video Diffusion Priors

This work addresses the challenge of streamed video depth estimation, which expects not only per-frame accuracy but, more importantly, cross-frame consistency. We argue that sharing contextual information between frames or clips is pivotal in fostering temporal consistency. Therefore, we reformulate depth prediction into a conditional generation problem to provide contextual information within a clip and across clips. Specifically, we propose a consistent context-aware training and inference strategy for arbitrarily long videos to provide cross-clip context. We sample independent noise levels for each frame within a clip during training while using a sliding window strategy and initializing overlapping frames with previously predicted frames without adding noise. Moreover, we design an effective training strategy to provide context within a clip. Extensive experimental results validate our design choices and demonstrate the superiority of our approach, dubbed ChronoDepth. Project page: https://xdimlab.github.io/ChronoDepth/.

cs.CV↗

arXiv2025

Multi-Instrument Search for Gamma-Ray Counterpart of X-ray Transients detected by EP/WXT

As a soft X-ray imager with unprecedentedly large field of view, EP/WXT has detected many (fast) X-ray transients, whose nature is very intriguing. Whether there is gamma-ray counterpart for the X-ray transient provides important implications for its origin. Some of them have been reported to be associated with GRB, however, a systematic study on the gamma-ray emission of these X-ray transients is lacking. In this work, we implemented a comprehensive targeted search for gamma-ray counterparts to 63 X-ray transients reported by EP/WXT during its first year of operation, using the dedicated multiple-instrument search pipeline, ETJASMIN, with GECAM-B, GECAM-C, Fermi/GBM, and \textit{Insight}-HXMT data. We find that 14 out of 63 (22\%) EP/WXT X-ray transients have gamma-ray counterparts. For other transients, ETJASMIN pipeline provided upper limit of gamma-ray emission, which is more stringent than that given by individual instrument. Moreover, we investigated the properties of the X-ray transients and their gamma-ray counterparts, and explored the relation between the x-ray transient and gamma-ray counterpart.

astro-ph.HE↗

arXiv2025

GRU: Mitigating the Trade-off between Unlearning and Retention for LLMs

Large language model (LLM) unlearning has demonstrated its essential role in removing privacy and copyright-related responses, crucial for their legal and safe applications. However, the pursuit of complete unlearning often comes with substantial costs due to its compromises in their general functionality, leading to a notorious trade-off between unlearning and retention. It motivates this paper to explore enhanced unlearning schemes that can mitigate this trade-off. Specifically, we propose Gradient Rectified Unlearning (GRU), an improved framework that regulates the directions of gradient updates during the unlearning procedure such that their side impacts on other, unrelated responses can be minimized. GRU is easy and general to implement, demonstrating practical effectiveness across a variety of well-established unlearning benchmarks.

cs.LG↗

arXiv2025

Asymptotic behavior of complete conformal metric near singular boundary

The boundary behavior of the singular Yamabe problem has been extensively studied near sufficiently smooth boundaries, while less is known about the asymptotic behavior of solutions near singular boundaries. In this paper, we study the asymptotic behaviors of solutions to the singular Yamabe problem with negative constant scalar curvature near singular boundaries and derive the optimal estimates for the background metric which is not necessarily conformally flat. In particular, we prove that the solutions are well approximated by the solutions in tangent cones at singular points on the boundaries.

math.AP↗

arXiv2025

Exciton-polaritons in a monolayer semiconductor coupled to van der Waals dielectric nanoantennas on a metallic mirror

Polaritons in nanophotonic structures have attracted long-standing interest owing to their fundamental importance and potential for applications in nonlinear and quantum optics. Nanoantennas (NAs) made from high refractive index dielectrics offer a suitable platform for polariton physics thanks to the strongly confined optical Mie resonances and low optical losses in contrast to metallic NAs. However, Mie modes are mainly confined within the NA, making inefficient their coupling with excitons in materials deposited externally. Here, we overcome this limitation by using a high-refractive index van der Waals material WS$_2$, which allows straightforward fabrication of NAs on gold. The combination of a 27 nm tall WS$_2$ NA and a gold substrate enables strong modification of the Mie mode distribution and field enhancement inside and in the vicinity of the NA. This allows observation of room-temperature Mie-polaritons (with a Rabi splitting above 80 meV) arising from the strong coupling between Mie modes and the exciton in a monolayer WSe$_2$ placed on WS$_2$/gold NAs. We demonstrate strong nonlinearity of Mie-polaritons, one order of magnitude higher than for excitons in monolayer WSe$_2$ on gold. Our results highlight applicability of van der Waals materials for the realisation of hybrid dielectric-metallic nanophotonics for the study of the strong light-matter interaction.

physics.optics↗

arXiv2025

Topological Jackiw-Rebbi States in Photonic Van der Waals Heterostructures

Topological phenomena, first studied in solid state physics, have seen increased interest for applications in nanophotonics owing to highly controllable light confinement with inherent robustness to defects. Photonic crystals can be designed to host topologically protected interface states for directional light transport, localization and robust lasing via tuning of the bulk topological invariant. At the same time, van der Waals (vdW) materials, in both their monolayer and quasi-bulk forms, are emerging as exciting additions to the field of nanophotonics, with a range of unique optoelectronic properties and intrinsic adherence to any type of host material, allowing fabrication of complex multi-layer structures. We present here a 1D topological photonic platform made from stacked nanostructured and planar layers of quasi-bulk WS$_2$ to achieve Jackiw-Rebbi (JR) interface states between two topologically distinct gratings in the near-infrared range around 750 nm. Such states are measured in the far-field with angle-resolved reflectance contrast measurements, exhibiting linewidth of 10 meV and highly directional emission with an angular bandwidth of 8.0$^\circ$. Subsequent local mapping of the structure via sub-wavelength resolution scattering-type scanning near-field optical microscopy (s-SNOM) reveals strong spatial confinement of the JR state to the grating interface region. Finally, we couple in the JR state the photoluminescence of monolayer WSe$_2$ incorporated in a five-layer vdW grating heterostructure, giving rise to directional enhancement of the excitonic emission of up to 22 times that of uncoupled monolayer, thus demonstrating the potential of the topological interface states for highly directional light emission in addition to light scattering.

physics.optics↗

arXiv2025

StructFlowBench: A Structured Flow Benchmark for Multi-turn Instruction Following

Multi-turn instruction following capability constitutes a core competency of large language models (LLMs) in real-world applications. Existing evaluation benchmarks predominantly focus on fine-grained constraint satisfaction and domain-specific capability assessment, yet overlook the crucial structural dependencies between dialogue turns that distinguish multi-turn from single-turn interactions. These structural dependencies not only reflect user intent but also establish an essential second dimension for the instruction following evaluation beyond constraint satisfaction. To address this gap, we propose StructFlowBench, a multi-turn instruction following benchmark with structural flow modeling. The benchmark defines an innovative structural flow framework with six fundamental inter-turn relationships. These relationships introduce novel structural constraints for model evaluation and also serve as generation parameters for creating customized dialogue flows tailored to specific scenarios. Adopting established LLM-based automatic evaluation methodologies, we conduct systematic evaluations of 13 leading open-source and closed-source LLMs. Experimental results reveal significant deficiencies in current models' comprehension of multi-turn dialogue structures. The code is available at https://github.com/MLGroupJLU/StructFlowBench.

cs.CL↗

arXiv2025

Pessimism Principle Can Be Effective: Towards a Framework for Zero-Shot Transfer Reinforcement Learning

Transfer reinforcement learning aims to derive a near-optimal policy for a target environment with limited data by leveraging abundant data from related source domains. However, it faces two key challenges: the lack of performance guarantees for the transferred policy, which can lead to undesired actions, and the risk of negative transfer when multiple source domains are involved. We propose a novel framework based on the pessimism principle, which constructs and optimizes a conservative estimation of the target domain's performance. Our framework effectively addresses the two challenges by providing an optimized lower bound on target performance, ensuring safe and reliable decisions, and by exhibiting monotonic improvement with respect to the quality of the source domains, thereby avoiding negative transfer. We construct two types of conservative estimations, rigorously characterize their effectiveness, and develop efficient distributed algorithms with convergence guarantees. Our framework provides a theoretically sound and practically robust solution for transfer learning in reinforcement learning.

cs.LG↗

arXiv2025

Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training

Mixture-of-Experts (MoE) architectures within Large Reasoning Models (LRMs) have achieved impressive reasoning capabilities by selectively activating experts to facilitate structured cognitive processes. Despite notable advances, existing reasoning models often suffer from cognitive inefficiencies like overthinking and underthinking. To address these limitations, we introduce a novel inference-time steering methodology called Reinforcing Cognitive Experts (RICE), designed to improve reasoning performance without additional training or complex heuristics. Leveraging normalized Pointwise Mutual Information (nPMI), we systematically identify specialized experts, termed ''cognitive experts'' that orchestrate meta-level reasoning operations characterized by tokens like '' ''. Empirical evaluations with leading MoE-based LRMs (DeepSeek-R1 and Qwen3-235B) on rigorous quantitative and scientific reasoning benchmarks demonstrate noticeable and consistent improvements in reasoning accuracy, cognitive efficiency, and cross-domain generalization. Crucially, our lightweight approach substantially outperforms prevalent reasoning-steering techniques, such as prompt design and decoding constraints, while preserving the model's general instruction-following skills. These results highlight reinforcing cognitive experts as a promising, practical, and interpretable direction to enhance cognitive efficiency within advanced reasoning models.

cs.AI↗

arXiv2025

Pre-training Generative Recommender with Multi-Identifier Item Tokenization

Generative recommendation autoregressively generates item identifiers to recommend potential items. Existing methods typically adopt a one-to-one mapping strategy, where each item is represented by a single identifier. However, this scheme poses issues, such as suboptimal semantic modeling for low-frequency items and limited diversity in token sequence data. To overcome these limitations, we propose MTGRec, which leverages Multi-identifier item Tokenization to augment token sequence data for Generative Recommender pre-training. Our approach involves two key innovations: multi-identifier item tokenization and curriculum recommender pre-training. For multi-identifier item tokenization, we leverage the RQ-VAE as the tokenizer backbone and treat model checkpoints from adjacent training epochs as semantically relevant tokenizers. This allows each item to be associated with multiple identifiers, enabling a single user interaction sequence to be converted into several token sequences as different data groups. For curriculum recommender pre-training, we introduce a curriculum learning scheme guided by data influence estimation, dynamically adjusting the sampling probability of each data group during recommender pre-training. After pre-training, we fine-tune the model using a single tokenizer to ensure accurate item identification for recommendation. Extensive experiments on three public benchmark datasets demonstrate that MTGRec significantly outperforms both traditional and generative recommendation baselines in terms of effectiveness and scalability.

cs.IR↗

arXiv2025

Deployable 3D mesoscale structures through wafer fabrication, geometric frustration and bistable auxeticity

Transforming planar mesoscale devices into precise 3-D architectures is vital for next-generation flexible electronics, implants, and adaptive optics, yet wafer-based manufacturing to free-standing 3-D structures remain elusive. We fabricate polyimide architected 2-D precursors whose bistable unit cells deploy into stable 3-D mesoscale structures. Target Gaussian curvature is encoded by conformally flattening the desired mesh and locally tuning each cell so its second equilibrium matches the required scaling factor, aided by a computed library of isotropically expanding, bistable microstructures. The resulting heterogeneous tessellations uniquely morph into complex shapes. A flat disk deploys into a hemispherical dome with sub-millimeter accuracy and retains its shape after indentation. The same process yields positive- and negative-curvature geometries and tunable-focus paraboloidal mirrors whose reflected laser patterns coincide with geometric optics calculations. Our wafer-compatible, generative algorithm extends far beyond flexible substrates, enabling truly deployable, high-performance electronics and optical devices.

physics.app-ph↗

arXiv2025

A generalized Riemann problem solver for a hyperbolic model of two-layer thin film flow

In this paper, a second-order generalized Riemann problem (GRP) solver is developed for a two-layer thin film model. Extending the first-order Godunov approach, the solver is used to construct a temporal-spatial coupled second-order GRP-based finite-volume method. Numerical experiments including comparisons to MUSCL finite-volume schemes with Runge-Kutta time stepping confirm the accuracy, efficiency and robustness of the higher-order ansatz. The construction of GRP methods requires to compute temporal derivatives of intermediate states in the entropy solution of the generalized Riemann problem. These derivatives are obtained from the Rankine-Hugoniot conditions as well as a characteristic decomposition using Riemann invariants. Notably, the latter can be computed explicitly for the two-layer thin film model, which renders this system to be very suitable for the GRP approach. Moreover, it becomes possible to determine the derivatives in an explicit, computationally cheap way.

math.NA↗

arXiv2025

DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning

Reinforcement learning (RL) with large language models shows promise in complex reasoning. However, its progress is hindered by the lack of large-scale training data that is sufficiently challenging, contamination-free and verifiable. To this end, we introduce DeepMath-103K, a large-scale mathematical dataset designed with high difficulty (primarily levels 5-9), rigorous decontamination against numerous benchmarks, and verifiable answers for rule-based RL reward. It further includes three distinct R1 solutions adaptable for diverse training paradigms such as supervised fine-tuning (SFT). Spanning a wide range of mathematical topics, DeepMath-103K fosters the development of generalizable and advancing reasoning. Notably, models trained on DeepMath-103K achieve state-of-the-art results on challenging mathematical benchmarks and demonstrate generalization beyond math such as biology, physics and chemistry, underscoring its broad efficacy. Data: https://huggingface.co/datasets/zwhe99/DeepMath-103K.

cs.CL↗