arXiv subjects

Ke Wang

Publications and source records attributed to Ke Wang.

At least 181 records · Page 10Linked to original sources

arXiv2025

Non-Markovian Noise Mitigation: Practical Implementation, Error Analysis, and the Role of Environment Spectral Properties

Quantum error mitigation(QEM), an error suppression strategy without the need for additional ancilla qubits for noisy intermediate-scale quantum~(NISQ) devices, presents a promising avenue for realizing quantum speedups of quantum computing algorithms on current quantum devices. However, prior investigations have predominantly been focused on Markovian noise. In this paper, we propose a non-Markovian Noise Mitigation(NMNM) method by extending the probabilistic error cancellation (PEC) method in the QEM framework to treat non-Markovian noise. We present the derivation of a time-local quantum master equation where the decoherence coefficients are directly obtained from bath correlation functions(BCFs), key properties of a non-Markovian environment that will make the error mitigation algorithms environment-aware. We further establish a direct connection between the overall approximation error and sampling overhead of QEM and the spectral property of the environment. Numerical simulations performed on a spin-boson model further validate the efficacy of our approach.

quant-ph↗

arXiv2025

Examining the Spin Structure of Altermagnetic Candidate MnTe Grown with Near Ideal Stoichiometry

Altermagnets are a recently-discovered class of materials with magnetic ordering that have a zero net magnetization and a momentum-dependent spin splitting in their band structure, arising from a collinear spin arrangement with alternating polarizations in the crystal lattice. The nickeline-structured manganese telluride (α-MnTe) is an attractive altermagnet candidate due to its predicted large spin splitting energy and a transition temperature near 300K. In this work, we present a thorough investigation of the spin structure of α-MnTe thin films grown by molecular beam epitaxy with very high crystal quality and low residual magnetization. The epitaxial α-MnTe films have a full-width-at-half-maximum of 0.1° as measured by x-ray-diffraction rocking curves and a root-mean-square roughness below 1 nm. Neutron diffraction measurements confirm the antiferromagnetic order in the α-MnTe film and show a Néel temperature of 307 K. Polarized neutron reflectometry detects a vanishingly small net magnetization which may be confined to the MnTe/InP interface, highlighting the near-ideal stoichiometry in the sample. In vacuo angle resolved photoemission spectroscopy reveals that the bulk band spectrum of the MnTe films is consistent with the weak altermagnetic order as theoretically predicted and observed for the high symmetry nodal plane in the center of the Brillouin zone. This study establishes optimized growth conditions for the synthesis of stoichiometric α-MnTe thin films which exhibit exceptional structural and magnetic ordering, thereby providing a robust platform for the precise characterization of their altermagnetic properties.

cond-mat.mtrl-sci↗

arXiv2025

MathCanvas: Intrinsic Visual Chain-of-Thought for Multimodal Mathematical Reasoning

While Large Language Models (LLMs) have excelled in textual reasoning, they struggle with mathematical domains like geometry that intrinsically rely on visual aids. Existing approaches to Visual Chain-of-Thought (VCoT) are often limited by rigid external tools or fail to generate the high-fidelity, strategically-timed diagrams necessary for complex problem-solving. To bridge this gap, we introduce MathCanvas, a comprehensive framework designed to endow unified Large Multimodal Models (LMMs) with intrinsic VCoT capabilities for mathematics. Our approach consists of two phases. First, a Visual Manipulation stage pre-trains the model on a novel 15.2M-pair corpus, comprising 10M caption-to-diagram pairs (MathCanvas-Imagen) and 5.2M step-by-step editing trajectories (MathCanvas-Edit), to master diagram generation and editing. Second, a Strategic Visual-Aided Reasoning stage fine-tunes the model on MathCanvas-Instruct, a new 219K-example dataset of interleaved visual-textual reasoning paths, teaching it when and how to leverage visual aids. To facilitate rigorous evaluation, we introduce MathCanvas-Bench, a challenging benchmark with 3K problems that require models to produce interleaved visual-textual solutions. Our model, BAGEL-Canvas, trained under this framework, achieves an 86% relative improvement over strong LMM baselines on MathCanvas-Bench, demonstrating excellent generalization to other public math benchmarks. Our work provides a complete toolkit-framework, datasets, and benchmark-to unlock complex, human-like visual-aided reasoning in LMMs. Project Page: https://mathcanvas.github.io/

cs.CV↗

arXiv2025

CodePlot-CoT: Mathematical Visual Reasoning by Thinking with Code-Driven Images

Recent advances in Large Language Models (LLMs) and Vision Language Models (VLMs) have shown significant progress in mathematical reasoning, yet they still face a critical bottleneck with problems requiring visual assistance, such as drawing auxiliary lines or plotting functions to solve the problems. Most LLMs and VLMs are constrained to text-only reasoning chains, while multimodal unified models that can generate interleaved text and images lack the necessary precision and controllability for such tasks. To address this, we propose CodePlot-CoT, a code-driven Chain-of-Thought paradigm for "thinking with images" in mathematics. Our approach leverages the VLM to generate text reasoning as well as executable plotting code, which is then rendered into images as "visual thought", to solve mathematical problems. To achieve this, we first construct Math-VR, the first large-scale, bilingual dataset and benchmark for Mathematics problems with Visual Reasoning, comprising 178K samples. Second, to create high-quality training data, we develop a state-of-the-art image-to-code converter specialized for parsing complex mathematical figures into codes. Finally, using these training data, we train the CodePlot-CoT model for solving mathematical problems. Experimental results show that our model achieves up to 21% increase over base model on our new benchmark, fully validating the efficacy of our proposed code-driven reasoning paradigm. Our work opens a new direction for multimodal mathematical reasoning and provides the community with the first large-scale dataset, comprehensive benchmark, and strong approach for such problems. To facilitate future research, we make our datasets, code, and pretrained models publicly available at https://github.com/HKU-MMLab/Math-VR-CodePlot-CoT.

cs.CV↗

arXiv2025

Meson-baryon scattering lengths without annihilation diagrams to order $p^4$ in heavy baryon chiral perturbation theory

We calculate the threshold $T$ matrices of the meson and baryon processes that have no annihilation diagrams: $π^{+}Σ^{+}$, $π^{+}Ξ^0$, $K^+p$, $K^+n$, and $\bar{K}^0Ξ^0$ to the fourth order in heavy baryon chiral perturbation theory. By performing least squares and Bayesian fits to the non-physical lattice QCD data, we determine the low-energy constants through both perturbative and non-perturbative iterative methods. By using these low-energy constants, we obtain the physical scattering lengths in these fits. The convergence behavior is not good across all channels in the perturbative method. The scattering lengths for the five channels, obtained by taking the median values from four different fitting approaches, are $a_{π^+Σ^+}=-0.16\pm 0.07\,\text{fm}$, $a_{π^+Ξ^0}=-0.04\pm0.04\,\text{fm}$, $a_{K^+p}=-0.41\pm 0.11\,\text{fm}$, $a_{K^+n}=-0.19\pm 0.10\,\text{fm}$, and $a_{\bar{K}^0Ξ^0}=-0.30\pm 0.07\,\text{fm}$, where the uncertainties are conservatively estimated by taking the maximum deviation between the median and extreme values of the statistical errors.

hep-ph↗

arXiv2025

AutoPrune: Each Complexity Deserves a Pruning Policy

The established redundancy in visual tokens within large vision-language models allows pruning to effectively reduce their substantial computational demands. Previous methods typically employ heuristic layer-specific pruning strategies where, although the number of tokens removed may differ across decoder layers, the overall pruning schedule is fixed and applied uniformly to all input samples and tasks, failing to align token elimination with the model's holistic reasoning trajectory. Cognitive science indicates that human visual processing often begins with broad exploration to accumulate evidence before narrowing focus as the target becomes distinct. Our experiments reveal an analogous pattern in these models. This observation suggests that neither a fixed pruning schedule nor a heuristic layer-wise strategy can optimally accommodate the diverse complexities inherent in different inputs. To overcome this limitation, we introduce Complexity-Adaptive Pruning (AutoPrune), a training-free, plug-and-play framework that tailors pruning policies to varying sample and task complexities. Specifically, AutoPrune quantifies the mutual information between visual and textual tokens, then projects this signal to a budget-constrained logistic retention curve. Each such logistic curve, defined by its unique shape, corresponds to the specific complexity of different tasks and can guarantee adherence to predefined computational constraints. We evaluate AutoPrune on standard vision-language tasks and on Vision-Language-Action models for autonomous driving. Notably, when applied to LLaVA-1.5-7B, our method prunes 89% of visual tokens and reduces inference FLOPs by 76.8% while retaining 96.7% of the original accuracy averaged over all tasks. This corresponds to a 9.1% improvement over the recent work PDrop, demonstrating the effectiveness. Code is available at https://github.com/AutoLab-SAI-SJTU/AutoPrune.

cs.CV↗

arXiv2025

Energy extraction from the accelerating Kerr black hole via magnetic reconnection in the plunging region and circular orbit region

Based on the magnetic reconnection mechanism, this study investigates how to extract energy effectively from an accelerating Kerr black hole in the plunging region and circular orbit region. After introducing the properties of accelerating black holes, including the event horizon, ergosphere, circular orbits, and innermost stable circular orbit, we investigate the magnetic reconnection process in the plunging region. Specifically, we analyze variations of the azimuthal angle with respect to the acceleration, examine changes in energy per enthalpy of decelerated plasma, and plot energy extraction efficiency along with permissible energy extraction regions. Results show that in the plunging region, at larger radii of reconnection locations, the accelerating black hole exhibits higher energy extraction efficiency than a Kerr black hole. Away from extremality, the acceleration parameter impedes energy extraction, while near extremality, it enhances extraction. We also study energy extraction in circular orbit region by plotting energy extraction efficiency within permissible regions. We find that the permissible energy extraction area is reduced and the efficiency exceeds that of Kerr black holes due to the existence of acceleration parameter. Larger acceleration parameters yield more effective energy extraction regardless of extremality, which is different from that in the plunging region. Additionally, energy extraction efficiency in the plunging region surpasses that in the circular orbit region, aligning with prior conclusions.

gr-qc↗

arXiv2025

A note on the improved sparse Hanson-Wright inequalities

We establish sparse Hanson-Wright inequalities for quadratic forms of sparse $α$-sub-exponential random vectors with exponent parameter $α\in(0, 2]$. In the regime $0< α\le 1$ we derive a refined inequality that is optimal in several canonical models. These results extend the classical Hanson-Wright bound to the sparse setting. Illustrative applications include covariance matrix estimation with incomplete observations, low-rank matrix approximation under the maximum norm with sparsified sketches, and concentration inequalities for sparse $α$-sub-exponential random vectors.

math.PR↗

arXiv2025

Agentic Reinforcement Learning with Implicit Step Rewards

Large language models (LLMs) are increasingly developed as autonomous agents using reinforcement learning (agentic RL) that reason and act in interactive environments. However, sparse and sometimes unverifiable rewards make it extremely challenging to assign credit when training LLM agents that serve as a policy. Recent work attempts to integrate process supervision into RL but suffers from biased annotation, reward hacking, high-variance from overly fine-grained rewards or failtures when state overlap is rare. We therefore introduce implicit step rewards for agentic RL (iStar), a general credit-assignment strategy that integrates seamlessly with standard RL algorithms without relying on additional rollouts or explicit step labels. Particularly, we alternatively optimize an implicit process reward model (PRM) with the policy model to generate implicit step rewards via a trajectory-based DPO objective. Theoretical analysis shows that this learning objective produces a step-wise reward function. Then the implicit step rewards are used to compute step-level advantages, which are combined with trajectory (or episode)-level advantages for policy updates, creating a self-reinforcing training loop. We evaluate our method on three challenging agent benchmarks, including WebShop and VisualSokoban, as well as open-ended social interactions with unverifiable rewards in SOTOPIA. Crucially, iStar shows superior performance over frontier LLMs and strong RL baselines across domains, achieving state-of-the-art results with higher sample-efficiency and training stability. Further analysis also demonstrates efficient exploration by iStar with increased rewards in both step- and episode-level while maintaining fewer steps to achieve task success. Code will be available soon.

cs.CL↗

arXiv2025

WebGen-Agent: Enhancing Interactive Website Generation with Multi-Level Feedback and Step-Level Reinforcement Learning

Agent systems powered by large language models (LLMs) have demonstrated impressive performance on repository-level code-generation tasks. However, for tasks such as website codebase generation, which depend heavily on visual effects and user-interaction feedback, current code agents rely only on simple code execution for feedback and verification. This approach fails to capture the actual quality of the generated code. In this paper, we propose WebGen-Agent, a novel website-generation agent that leverages comprehensive and multi-level visual feedback to iteratively generate and refine the website codebase. Detailed and expressive text descriptions and suggestions regarding the screenshots and GUI-agent testing of the websites are generated by a visual language model (VLM), together with scores that quantify their quality. The screenshot and GUI-agent scores are further integrated with a backtracking and select-best mechanism, enhancing the performance of the agent. Utilizing the accurate visual scores inherent in the WebGen-Agent workflow, we further introduce \textit{Step-GRPO with Screenshot and GUI-agent Feedback} to improve the ability of LLMs to act as the reasoning engine of WebGen-Agent. By using the screenshot and GUI-agent scores at each step as the reward in Step-GRPO, we provide a dense and reliable process supervision signal, which effectively improves the model's website-generation ability. On the WebGen-Bench dataset, WebGen-Agent increases the accuracy of Claude-3.5-Sonnet from 26.4% to 51.9% and its appearance score from 3.0 to 3.9, outperforming the previous state-of-the-art agent system. Additionally, our Step-GRPO training approach increases the accuracy of Qwen2.5-Coder-7B-Instruct from 38.9% to 45.4% and raises the appearance score from 3.4 to 3.7.

cs.CL↗

arXiv2025

VoiceAssistant-Eval: Benchmarking AI Assistants across Listening, Speaking, and Viewing

The growing capabilities of large language models and multimodal systems have spurred interest in voice-first AI assistants, yet existing benchmarks are inadequate for evaluating the full range of these systems' capabilities. We introduce VoiceAssistant-Eval, a comprehensive benchmark designed to assess AI assistants across listening, speaking, and viewing. VoiceAssistant-Eval comprises 10,497 curated examples spanning 13 task categories. These tasks include natural sounds, music, and spoken dialogue for listening; multi-turn dialogue, role-play imitation, and various scenarios for speaking; and highly heterogeneous images for viewing. To demonstrate its utility, we evaluate 21 open-source models and GPT-4o-Audio, measuring the quality of the response content and speech, as well as their consistency. The results reveal three key findings: (1) proprietary models do not universally outperform open-source models; (2) most models excel at speaking tasks but lag in audio understanding; and (3) well-designed smaller models can rival much larger ones. Notably, the mid-sized Step-Audio-2-mini (7B) achieves more than double the listening accuracy of LLaMA-Omni2-32B-Bilingual. However, challenges remain: multimodal (audio plus visual) input and role-play voice imitation tasks are difficult for current models, and significant gaps persist in robustness and safety alignment. VoiceAssistant-Eval identifies these gaps and establishes a rigorous framework for evaluating and guiding the development of next-generation AI assistants. Code and data will be released at https://mathllm.github.io/VoiceAssistantEval/ .

cs.CL↗

arXiv2025

LM-Searcher: Cross-domain Neural Architecture Search with LLMs via Unified Numerical Encoding

Recent progress in Large Language Models (LLMs) has opened new avenues for solving complex optimization problems, including Neural Architecture Search (NAS). However, existing LLM-driven NAS approaches rely heavily on prompt engineering and domain-specific tuning, limiting their practicality and scalability across diverse tasks. In this work, we propose LM-Searcher, a novel framework that leverages LLMs for cross-domain neural architecture optimization without the need for extensive domain-specific adaptation. Central to our approach is NCode, a universal numerical string representation for neural architectures, which enables cross-domain architecture encoding and search. We also reformulate the NAS problem as a ranking task, training LLMs to select high-performing architectures from candidate pools using instruction-tuning samples derived from a novel pruning-based subspace sampling strategy. Our curated dataset, encompassing a wide range of architecture-performance pairs, encourages robust and transferable learning. Comprehensive experiments demonstrate that LM-Searcher achieves competitive performance in both in-domain (e.g., CNNs for image classification) and out-of-domain (e.g., LoRA configurations for segmentation and generation) tasks, establishing a new paradigm for flexible and generalizable LLM-based architecture search. The datasets and models will be released at https://github.com/Ashone3/LM-Searcher.

cs.CL↗

arXiv2025

SATBench: Benchmarking LLMs' Logical Reasoning via Automated Puzzle Generation from SAT Formulas

We introduce SATBench, a benchmark for evaluating the logical reasoning capabilities of large language models (LLMs) through logical puzzles derived from Boolean satisfiability (SAT) problems. Unlike prior work that focuses on inference rule-based reasoning, which often involves deducing conclusions from a set of premises, our approach leverages the search-based nature of SAT problems, where the objective is to find a solution that fulfills a specified set of logical constraints. Each instance in SATBench is generated from a SAT formula, then translated into a puzzle using LLMs. The generation process is fully automated and allows for adjustable difficulty by varying the number of clauses. All 2100 puzzles are validated through both LLM-based and solver-based consistency checks, with human validation on a subset. Experimental results show that even the strongest model, o4-mini, achieves only 65.0% accuracy on hard UNSAT problems, close to the random baseline of 50%. Our error analysis reveals systematic failures such as satisfiability bias, context inconsistency, and condition omission, highlighting limitations of current LLMs in search-based logical reasoning. Our code and data are publicly available at https://github.com/Anjiang-Wei/SATBench

cs.AI↗

arXiv2025

EquiBench: Benchmarking Large Language Models' Reasoning about Program Semantics via Equivalence Checking

As large language models (LLMs) become integral to code-related tasks, a central question emerges: Do LLMs truly understand program semantics? We introduce EquiBench, a new benchmark for evaluating LLMs through equivalence checking, i.e., determining whether two programs produce identical outputs for all possible inputs. Unlike prior code generation benchmarks, this task directly tests a model's ability to reason about program semantics. EquiBench consists of 2400 program pairs across four languages and six categories. These pairs are generated through program analysis, compiler scheduling, and superoptimization, ensuring high-confidence labels, nontrivial difficulty, and full automation. We evaluate 19 state-of-the-art LLMs and find that in the most challenging categories, the best accuracies are 63.8% and 76.2%, only modestly above the 50% random baseline. Further analysis reveals that models often rely on syntactic similarity rather than exhibiting robust reasoning about program semantics, highlighting current limitations. Our code and dataset are publicly available at https://github.com/Anjiang-Wei/equibench

cs.LG↗

arXiv2025

Fine-Tuning Large Multimodal Models for Automatic Pronunciation Assessment

Automatic Pronunciation Assessment (APA) is critical for Computer-Assisted Language Learning (CALL), requiring evaluation across multiple granularities and aspects. Large Multimodal Models (LMMs) present new opportunities for APA, but their effectiveness in fine-grained assessment remains uncertain. This work investigates fine-tuning LMMs for APA using the Speechocean762 dataset and a private corpus. Fine-tuning significantly outperforms zero-shot settings and achieves competitive results on single-granularity tasks compared to public and commercial systems. The model performs well at word and sentence levels, while phoneme-level assessment remains challenging. We also observe that the Pearson Correlation Coefficient (PCC) reaches 0.9, whereas Spearman's rank Correlation Coefficient (SCC) remains around 0.6, suggesting that SCC better reflects ordinal consistency. These findings highlight both the promise and limitations of LMMs for APA and point to future work on fine-grained modeling and rank-aware evaluation.

cs.CL↗

arXiv2025

The ALMA-QUARKS survey: Extensive detection of acetamide in multiple high-mass star-forming regions

Acetamide (CH$_{3}$CONH$_{2}$), a key interstellar amide and a methyl derivative of formamide (NH$_{2}$CHO), has been sparsely detected, limiting insights into its prebiotic relevance. We present the first systematic survey for acetamide toward 52 hot molecular cores using ALMA Band 6 data. Acetamide has been detected in 10 cores, markedly expanding the inventory of known emitters. The derived column densities of acetamide range from $(2.5\pm0.9)\times10^{14}$ to $(1.5\pm0.6)\times10^{16}$ cm$^{-2}$, compared to formamide's $(1.1\pm0.1)\times10^{15}$ to $(6.9\pm0.4)\times10^{16}$ cm$^{-2}$. The nearly constant abundance ratios (~3-9) and strong abundance correlation between the two amides across sources suggest a chemically linked formation pathway, likely on grain surfaces. The presence of peptide-like molecules in these regions implies that complex organic species can survive star formation processes, offering a potential pathway toward prebiotic chemistry. These findings constrain the dominant grain surface formation routes of acetamide, confirm its broader prevalence in highmass star-forming regions, and underscore the importance of targeted amide surveys in tracing the chemical evolution toward prebiotic complexity.

astro-ph.GA↗

arXiv2025

Scientific Objectives of the Xue-shan-mu-chang 15-meter Submillimeter Telescope

Submillimeter astronomy is poised to revolutionize our understanding of the Universe by revealing cosmic phenomena hidden from optical and near-infrared observations, particularly those associated with interstellar dust, molecular gas, and star formation. The Xue-shan-mu-chang 15-meter submillimeter telescope (XSMT-15m), to be constructed at a premier high-altitude site (4813 m) in Qinghai, China, marks a major milestone for Chinese astronomy, establishing the China mainland's first independently developed, world-class submillimeter facility. Equipped with state-of-the-art instruments, XSMT-15m will address a diverse range of frontier scientific questions spanning extragalactic astronomy, Galactic structure, time-domain astrophysics, and astrochemistry. In synergy with current and forthcoming observatories, XSMT-15m will illuminate the formation and evolution of galaxies, unravel the physical and chemical processes shaping the interstellar medium, and explore transient phenomena in the submillimeter regime. These capabilities will advance our understanding across extragalactic astronomy, Galactic ecology, astrochemistry, and time-domain astrophysics, inaugurating a new era for submillimeter research in China and the northern hemisphere.

astro-ph.GA↗

arXiv2025

Dense Molecular Ring-like structure in gaseous CO depletion region G34.74-0.12

We report the discovery of a dense molecular ring-like structure in a dense (10$^5$ cm$^{-3}$), cold (pc-scale CO depletion at a factor of 5), and young (10$^4$ year) star-forming region G34.74-0.12, revealed by C$^{18}$O (2-1), HNC (1-0), and N$_2$H$^+$ (1-0) observations with the Atacama Large Millimeter/submillimeter Array (ALMA). The ring-like structure is redshifted with respect to the clump, spanning from $V_{\rm sys,lsr} + 0.9$ to $V_{\rm sys,lsr} + 2.9$ km s$^{-1}$, with a total mass of 109 $M_{\odot}$. It is spatially coincident with 1.3 mm and 3.0 mm dust continuum emission from cores, and several protostellar outflows. However, no free-free emission or H\textsc{ii} region is detected in association with this structure. With a slow expansion speed indicated by the position-velocity diagram, this ring structure differs from rings previously identified in more evolved star-forming regions. Possible explanations for the ring-like structure include a relic wind-blown bubble produced by a deeply embedded young stellar object, a hollow cavity formed by cloud-cloud interactions, a gas ring resulting from a temperature gradient, or a line-of-sight superposition of multiple outflows or dense clouds. This discovery offers a rare observational glimpse into the earliest dynamical processes involved in massive star formation.

astro-ph.GA↗