arXiv subjects

Yi Zhu

Publications and source records attributed to Yi Zhu.

At least 73 records · Page 4Linked to original sources

arXiv2025

TrainVerify: Equivalence-Based Verification for Distributed LLM Training

Training large language models (LLMs) at scale requires parallel execution across thousands of devices, incurring enormous computational costs. Yet, these costly distributed trainings are rarely verified, leaving them prone to silent errors and potentially wasting millions of GPU hours. We introduce TrainVerify, a system for verifiable distributed training of LLMs. Given a deep learning model's logical specification as the ground truth, TrainVerify formally verifies that a distributed parallel execution plan is mathematically equivalent to it. Direct verification is notoriously difficult due to the sheer scale of LLMs which often involves billions of variables and highly intricate computation graphs. Therefore, TrainVerify introduces shape-reduction techniques and a stage-wise parallel verification algorithm that significantly reduces complexity while preserving formal correctness. TrainVerify scales to frontier LLMs, including the successful verification of the Llama3 (405B) and DeepSeek-V3 (671B) training plans.

cs.DC↗

arXiv2025

The impact of parameter spread of high-temperature superconducting Josephson junctions on the performance of quantum-based voltage standards

Quantum metrology based on Josephson junction array reproduces the most accurate desired voltage by far, therefore being introduced to provide voltage standards worldwide. In this work, we quantitatively analyzed the dependence of the first Shapiro step height of the junction array at 50 GHz on the parameter spread of 10,000 Josephson junctions by numerical simulation with resistively shunted junction model. The results indicate an upper limit spread of the critical current and resistance of the Josephson junctions. Specifically, to keep the maximum first Shapiro step above 0.88 mA, the critical current standard deviation, $σ$, should not exceed 25%, and for it to stay above 0.6 mA, the resistance standard deviation should not exceed 1.5%.

cond-mat.supr-con↗

arXiv2025

Enhancing Efficiency and Propulsion in Bio-mimetic Robotic Fish through End-to-End Deep Reinforcement Learning

Aquatic organisms are known for their ability to generate efficient propulsion with low energy expenditure. While existing research has sought to leverage bio-inspired structures to reduce energy costs in underwater robotics, the crucial role of control policies in enhancing efficiency has often been overlooked. In this study, we optimize the motion of a bio-mimetic robotic fish using deep reinforcement learning (DRL) to maximize propulsion efficiency and minimize energy consumption. Our novel DRL approach incorporates extended pressure perception, a transformer model processing sequences of observations, and a policy transfer scheme. Notably, significantly improved training stability and speed within our approach allow for end-to-end training of the robotic fish. This enables agiler responses to hydrodynamic environments and possesses greater optimization potential compared to pre-defined motion pattern controls. Our experiments are conducted on a serially connected rigid robotic fish in a free stream with a Reynolds number of 6000 using computational fluid dynamics (CFD) simulations. The DRL-trained policies yield impressive results, demonstrating both high efficiency and propulsion. The policies also showcase the agent's embodiment, skillfully utilizing its body structure and engaging with surrounding fluid dynamics, as revealed through flow analysis. This study provides valuable insights into the bio-mimetic underwater robots optimization through DRL training, capitalizing on their structural advantages, and ultimately contributing to more efficient underwater propulsion systems.

cs.RO↗

arXiv2025

CAP-Net: A Unified Network for 6D Pose and Size Estimation of Categorical Articulated Parts from a Single RGB-D Image

This paper tackles category-level pose estimation of articulated objects in robotic manipulation tasks and introduces a new benchmark dataset. While recent methods estimate part poses and sizes at the category level, they often rely on geometric cues and complex multi-stage pipelines that first segment parts from the point cloud, followed by Normalized Part Coordinate Space (NPCS) estimation for 6D poses. These approaches overlook dense semantic cues from RGB images, leading to suboptimal accuracy, particularly for objects with small parts. To address these limitations, we propose a single-stage Network, CAP-Net, for estimating the 6D poses and sizes of Categorical Articulated Parts. This method combines RGB-D features to generate instance segmentation and NPCS representations for each part in an end-to-end manner. CAP-Net uses a unified network to simultaneously predict point-wise class labels, centroid offsets, and NPCS maps. A clustering algorithm then groups points of the same predicted class based on their estimated centroid distances to isolate each part. Finally, the NPCS region of each part is aligned with the point cloud to recover its final pose and size. To bridge the sim-to-real domain gap, we introduce the RGBD-Art dataset, the largest RGB-D articulated dataset to date, featuring photorealistic RGB images and depth noise simulated from real sensors. Experimental evaluations on the RGBD-Art dataset demonstrate that our method significantly outperforms the state-of-the-art approach. Real-world deployments of our model in robotic tasks underscore its robustness and exceptional sim-to-real transfer capabilities, confirming its substantial practical utility. Our dataset, code and pre-trained models are available on the project page.

cs.CV↗

arXiv2025

rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Dataset

Advancing code reasoning in large language models (LLMs) is fundamentally limited by the scarcity of high-difficulty datasets, especially those with verifiable input-output test cases necessary for rigorous solution validation at scale. We introduce rStar-Coder, which significantly improves LLM code reasoning capabilities by constructing a large-scale, verified dataset of 418K competition-level code problems, 580K long-reasoning solutions along with rich test cases of varying difficulty. This is achieved through three core contributions: (1) we curate competitive programming code problems and oracle solutions to synthesize new, solvable problems; (2) we introduce a reliable input-output test case synthesis pipeline that decouples the generation into a three-step input generation method and a mutual verification mechanism for effective output labeling; (3) we augment problems with high-quality, test-case-verified long-reasoning solutions. Extensive experiments on Qwen models (1.5B-14B) across various code reasoning benchmarks demonstrate the superiority of rStar-Coder dataset, achieving leading performance comparable to frontier reasoning LLMs with much smaller model sizes. On LiveCodeBench, rStar-Coder improves Qwen2.5-7B from 17.4% to an impressive 57.3%, and Qwen2.5-14B from 23.3% to 62.5%, surpassing o3-mini (low) by3.1%. On the more challenging USA Computing Olympiad, our 7B model achieves an average pass@1 accuracy of 16.15%, outperforming the frontier-level QWQ-32B. Code and the dataset will be released at https://github.com/microsoft/rStar.

cs.CL↗

arXiv2025

Global well-posedness of 2D incompressible MHD equations without magnetic diffusion

In recent years, the global existence of classical solutions to the Cauchy problem for 2D incompressible viscous MHD equations without magnetic diffusion has been proved in \cite{Ren,TZhang}, under the assumption that initial data is close to equilibrium states with nontrivial magnetic field, and the perturbation is small in some suitable spaces, say for instance, the Sobolev spaces with negative exponents. It leads to an interesting open question: Can one establish the global existence of classical solutions without the extra help from Sobolev spaces with negative exponents like its counterparts of ideal MHD ( i.e. without viscosity and magnetic diffusion), and fully dissipative MHD (i.e. with both viscosity and magnetic diffusion)? This paper offers an affirmative answer to this question. In fact, we will establish the existence of a global unique solution for initial perturbations being small in $H^2(\mathbb{R}^2)$. The key idea is further exploring the structure of system, using dispersive effects of Alfvén waves in the direction which is transversal to the dissipation favorable direction. This motivates our key strategy to treat the wildest nonlinear terms as an artificial linear term. These observations help us to construct some interesting quantities which improves the nonlinearity order for the wildest terms, and to control them by terms with better properties.

math.AP↗

arXiv2025

DGRO: Enhancing LLM Reasoning via Exploration-Exploitation Control and Reward Variance Management

Inference scaling further accelerates Large Language Models (LLMs) toward Artificial General Intelligence (AGI), with large-scale Reinforcement Learning (RL) to unleash long Chain-of-Thought reasoning. Most contemporary reasoning approaches usually rely on handcrafted rule-based reward functions. However, the tarde-offs of exploration and exploitation in RL algorithms involves multiple complex considerations, and the theoretical and empirical impacts of manually designed reward functions remain insufficiently explored. In this paper, we propose Decoupled Group Reward Optimization (DGRO), a general RL algorithm for LLM reasoning. On the one hand, DGRO decouples the traditional regularization coefficient into two independent hyperparameters: one scales the policy gradient term, and the other regulates the distance from the sampling policy. This decoupling not only enables precise control over balancing exploration and exploitation, but also can be seamlessly extended to Online Policy Mirror Descent (OPMD) algorithms in Kimi k1.5 and Direct Reward Optimization. On the other hand, we observe that reward variance significantly affects both convergence speed and final model performance. We conduct both theoretical analysis and extensive empirical validation to assess DGRO, including a detailed ablation study that investigates its performance and optimization dynamics. Experimental results show that DGRO achieves state-of-the-art performance on the Logic dataset with an average accuracy of 96.9\%, and demonstrates strong generalization across mathematical benchmarks.

cs.LG↗

arXiv2025

Clickbait Detection via Large Language Models

Clickbait, which aims to induce users with some surprising and even thrilling headlines for increasing click-through rates, permeates almost all online content publishers, such as news portals and social media. Recently, Large Language Models (LLMs) have emerged as a powerful instrument and achieved tremendous success in a series of NLP downstream tasks. However, it is not yet known whether LLMs can be served as a high-quality clickbait detection system. In this paper, we analyze the performance of LLMs in the few-shot and zero-shot scenarios on several English and Chinese benchmark datasets. Experimental results show that LLMs cannot achieve the best results compared to the state-of-the-art deep and fine-tuning PLMs methods. Different from human intuition, the experiments demonstrated that LLMs cannot make satisfied clickbait detection just by the headlines.

cs.CL↗

arXiv2025

DeltaDPD: Exploiting Dynamic Temporal Sparsity in Recurrent Neural Networks for Energy-Efficient Wideband Digital Predistortion

Digital Predistortion (DPD) is a popular technique to enhance signal quality in wideband RF power amplifiers (PAs). With increasing bandwidth and data rates, DPD faces significant energy consumption challenges during deployment, contrasting with its efficiency goals. State-of-the-art DPD models rely on recurrent neural networks (RNN), whose computational complexity hinders system efficiency. This paper introduces DeltaDPD, exploring the dynamic temporal sparsity of input signals and neuronal hidden states in RNNs for energy-efficient DPD, reducing arithmetic operations and memory accesses while preserving satisfactory linearization performance. Applying a TM3.1a 200MHz-BW 256-QAM OFDM signal to a 3.5 GHz GaN Doherty RF PA, DeltaDPD achieves -50.03 dBc in Adjacent Channel Power Ratio (ACPR), -37.22 dB in Normalized Mean Square Error (NMSE) and -38.52 dBc in Error Vector Magnitude (EVM) with 52% temporal sparsity, leading to a 1.8X reduction in estimated inference power. The DeltaDPD code will be released after formal publication at https://www.opendpd.com.

eess.SP↗

arXiv2025

A thermodynamics-based turbulence model for isothermal compressible flows

This study presents a new turbulence model for isothermal compressible flows. The model is derived by combining the Favre averaging and the Conservation-dissipation formalism -- a newly developed thermodynamics theory. The latter provides a systematic methodology to construct closure relations that intrinsically satisfy the first and second laws of thermodynamics. The new model is a hyperbolic system of first-order partial differential equations. It has a number of numerical advantages, and addresses some drawbacks of classical turbulence models by resolving the non-physical infinite information propagation paradox of the parabolic-type models and accurately capturing the interaction between compressibility and turbulence dissipation. Furthermore, we show the compatibility of the proposed model with Prandtl's one-equation model for incompressible flows by deliberately rescaling the model and studying its low Mach number limit.

math.AP↗

arXiv2025

Multi-Agent Relative Investment Games in a Jump Diffusion Market with Deep Reinforcement Learning Algorithm

This paper focuses on multi-agent stochastic differential games for jump-diffusion systems. On one hand, we study the multi-agent game for optimal investment in a jump-diffusion market. We derive constant Nash equilibria and provide sufficient conditions for their existence and uniqueness for exponential, power, and logarithmic utilities, respectively. On the other hand, we introduce a computational framework based on the actor-critic method in deep reinforcement learning to solve the stochastic control problem with jumps. We extend this algorithm to address the multi-agent game with jumps and utilize parallel computing to enhance computational efficiency. We present numerical examples of the Merton problem with jumps, linear quadratic regulators, and the optimal investment game under various settings to demonstrate the accuracy, efficiency, and robustness of the proposed method. In particular, neural network solutions numerically converge to the derived constant Nash equilibrium for the multi-agent game.

math.OC↗

arXiv2025

Differential Transformer

Transformer tends to overallocate attention to irrelevant context. In this work, we introduce Diff Transformer, which amplifies attention to the relevant context while canceling noise. Specifically, the differential attention mechanism calculates attention scores as the difference between two separate softmax attention maps. The subtraction cancels noise, promoting the emergence of sparse attention patterns. Experimental results on language modeling show that Diff Transformer outperforms Transformer in various settings of scaling up model size and training tokens. More intriguingly, it offers notable advantages in practical applications, such as long-context modeling, key information retrieval, hallucination mitigation, in-context learning, and reduction of activation outliers. By being less distracted by irrelevant context, Diff Transformer can mitigate hallucination in question answering and text summarization. For in-context learning, Diff Transformer not only enhances accuracy but is also more robust to order permutation, which was considered as a chronic robustness issue. The results position Diff Transformer as a highly effective and promising architecture to advance large language models.

cs.CL↗

arXiv2025

EMOVA: Empowering Language Models to See, Hear and Speak with Vivid Emotions

GPT-4o, an omni-modal model that enables vocal conversations with diverse emotions and tones, marks a milestone for omni-modal foundation models. However, empowering Large Language Models to perceive and generate images, texts, and speeches end-to-end with publicly available data remains challenging for the open-source community. Existing vision-language models rely on external tools for speech processing, while speech-language models still suffer from limited or totally without vision-understanding capabilities. To address this gap, we propose the EMOVA (EMotionally Omni-present Voice Assistant), to enable Large Language Models with end-to-end speech abilities while maintaining the leading vision-language performance. With a semantic-acoustic disentangled speech tokenizer, we surprisingly notice that omni-modal alignment can further enhance vision-language and speech abilities compared with the bi-modal aligned counterparts. Moreover, a lightweight style module is introduced for the flexible speech style controls including emotions and pitches. For the first time, EMOVA achieves state-of-the-art performance on both the vision-language and speech benchmarks, and meanwhile, supporting omni-modal spoken dialogue with vivid emotions.

cs.CV↗

arXiv2025

Edge states in super honeycomb structures with PT-symmetric deformations

The existence of edge states is one of the most vital properties of topological insulators. Although tremendous success has been accomplished in describing and explaining edge states associated with PT symmetry breaking, little work has been done on PT symmetry preserving cases. Two-dimensional Schroedinger operators with super honeycomb lattice potentials always have double Dirac cones at the Gamma point - the zero momentum point on their energy bands due to C6 symmetry, PT symmetry, and the "folding" symmetry - caused by an additional translation symmetry. There are two topologically different ways to deform such a system by PT symmetry preserving but folding symmetry breaking perturbations. Interestingly, there exist two gapped edge states on the interface between such two kinds of perturbed materials. In this paper, we illustrate the existence of such PT preserving edge states rigorously for the first time. We use a domain wall modulated Schroedinger operator to model the phenomenon under small perturbations and rigorously prove the existence of two gapped edge states. We also provide a brief interpretation from the point of view of "topology" by the parities of degenerate bulk modes. Our work thoroughly explains the existence of "helical" like edge states in super honeycomb configurations and lays a foundation for the descriptions of topologies of such systems.

math-ph↗

arXiv2025

Reconstruction of dynamical systems from data without time labels

In this paper, we study the method to reconstruct dynamical systems from data without time labels. Data without time labels appear in many applications, such as molecular dynamics, single-cell RNA sequencing etc. Reconstruction of dynamical system from time sequence data has been studied extensively. However, these methods do not apply if time labels are unknown. Without time labels, sequence data becomes distribution data. Based on this observation, we propose to treat the data as samples from a probability distribution and try to reconstruct the underlying dynamical system by minimizing the distribution loss, sliced Wasserstein distance more specifically. Extensive experiment results demonstrate the effectiveness of the proposed method.

cs.LG↗

arXiv2025

Redefining Simplicity: Benchmarking Large Language Models from Lexical to Document Simplification

Text simplification (TS) refers to the process of reducing the complexity of a text while retaining its original meaning and key information. Existing work only shows that large language models (LLMs) have outperformed supervised non-LLM-based methods on sentence simplification. This study offers the first comprehensive analysis of LLM performance across four TS tasks: lexical, syntactic, sentence, and document simplification. We compare lightweight, closed-source and open-source LLMs against traditional non-LLM methods using automatic metrics and human evaluations. Our experiments reveal that LLMs not only outperform non-LLM approaches in all four tasks but also often generate outputs that exceed the quality of existing human-annotated references. Finally, we present some future directions of TS in the era of LLMs.

cs.CL↗

arXiv2025

RPGBENCH: Evaluating Large Language Models as Role-Playing Game Engines

We present RPGBench, the first benchmark designed to evaluate large language models (LLMs) as text-based role-playing game (RPG) engines. RPGBench comprises two core tasks: Game Creation (GC) and Game Simulation (GS). In GC, an LLM must craft a valid and playable RPG world using a structured event-state representation, ensuring logical coherence and proper termination conditions. In GS, the LLM simulates interactive gameplay across multiple rounds while consistently updating states and enforcing game rules. To comprehensively assess performance, RPGBench integrates objective and subjective evaluation methodologies. Objective measures verify adherence to event mechanics and check variable updates without requiring human intervention. Subjective measures, such as content interestingness, action quality, and role-playing capability, are evaluated via an LLM-as-a-judge framework, where a strong LLM grades each candidate's outputs. Empirical results demonstrate that state-of-the-art LLMs can produce engaging stories but often struggle to implement consistent, verifiable game mechanics, particularly in long or complex scenarios. By combining structured, rule-based assessments with LLM-based judgments, RPGBench provides a new standard for evaluating how well LLMs can balance creativity, coherence, and complexity in text-based RPGs, opening avenues for more immersive and controllable interactive storytelling.

cs.CL↗

arXiv2025

New Evaluation Paradigm for Lexical Simplification

Lexical Simplification (LS) methods use a three-step pipeline: complex word identification, substitute generation, and substitute ranking, each with separate evaluation datasets. We found large language models (LLMs) can simplify sentences directly with a single prompt, bypassing the traditional pipeline. However, existing LS datasets are not suitable for evaluating these LLM-generated simplified sentences, as they focus on providing substitutes for single complex words without identifying all complex words in a sentence. To address this gap, we propose a new annotation method for constructing an all-in-one LS dataset through human-machine collaboration. Automated methods generate a pool of potential substitutes, which human annotators then assess, suggesting additional alternatives as needed. Additionally, we explore LLM-based methods with single prompts, in-context learning, and chain-of-thought techniques. We introduce a multi-LLMs collaboration approach to simulate each step of the LS task. Experimental results demonstrate that LS based on multi-LLMs approaches significantly outperforms existing baselines.

cs.CL↗