arXiv subjects

Peng Cheng

Publications and source records attributed to Peng Cheng.

At least 73 records · Page 4Linked to original sources

arXiv2026

On the Cuspy Structure of Rotating Wormhole Shadows

We investigate the shadow cast by a rotating traversable wormhole in the Teo class endowed with a general redshift function, with particular emphasis on the emergence of cuspy structures. The shadow boundary is the common envelope of two critical orbit families: unstable circular orbits outside the throat and orbits at the throat itself. The formation of cusps, marking the transition between smooth and cuspy shadow boundaries, only becomes possible when the redshift parameter $λ$ is allowed to vary. Moreover, we uncover a universal critical value $λ_c$ that signals the onset of the cusp. A phase diagram characterized by the spin and redshift parameters reveals four distinct morphologies: smooth, cuspy, ears touching, and throat drowning. The morphology of the wormhole shadow may provide observational diagnostics for the different compact objects in future high-resolution imaging observations.

gr-qc↗

arXiv2026

Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-training

Continual pre-training on small-scale task-specific data is an effective method for improving large language models in new target fields, yet it risks catastrophic forgetting of their original capabilities. A common solution is to re-weight training data mixtures from source and target fields on a domain space to achieve balanced performance. Previous domain reweighting strategies rely on manual designation with certain heuristics based on human intuition or empirical results. In this work, we prove that more general heuristics can be parameterized by proposing Data Mixing Agent, the first model-based, end-to-end framework that learns to re-weight domains. The agent learns generalizable heuristics through reinforcement learning on large quantities of data mixing trajectories with corresponding feedback from an evaluation environment. Experiments in continual pre-training on math reasoning show that Data Mixing Agent outperforms strong baselines in achieving balanced performance across source and target field benchmarks. Furthermore, it generalizes well across unseen source fields, target models, and domain spaces without retraining. Direct application to the code generation field also indicates its adaptability across target domains. Further analysis showcases the agents' well-aligned heuristics with human intuitions and their efficiency in achieving superior model performance with less source-field data.

cs.LG↗

arXiv2026

Supercell-size scaling of moiré band flatness

In moiré superlattices, the band flatness governs the degree of wave localization, which is central to harnessing emergent phenomena and designing functional meta-devices. While research has focused on the magic conditions such as magic angle and magic distance for optimal flatness, a fundamental understanding of how flatness changes with the supercell size has remained elusive. Here, we establish a universal scaling between band flatness and supercell size. Theoretically, by recognizing the statistical equivalence between structural perturbations in moiré superlattices and disordered systems, we introduce the Thouless number to evaluate the strength of moiré localization. This approach allows us to establish a scaling theory for the evolution of band flatness with the supercell size, from which an analytical expression is derived. Our full-wave simulations with one-dimensional and two-dimensional moiré superlattices show excellent agreement with the theoretical prediction. Our work reveals a general scaling law for moiré band flatness, offering a new perspective for understanding and designing moiré-based resonant systems.

physics.optics↗

arXiv2026

LUMINA: LLM-Guided GPU Architecture Exploration via Bottleneck Analysis

GPU design space exploration (DSE) for modern AI workloads, such as Large-Language Model (LLM) inference, is challenging because of GPUs' vast, multi-modal design spaces, high simulation costs, and complex design optimization objectives (e.g. performance, power and area trade-offs). Existing automated DSE methods are often prohibitively expensive, either requiring an excessive number of exploration samples or depending on intricate, manually crafted analyses of interdependent critical paths guided by human heuristics. We present LUMINA, an LLM-driven GPU architecture exploration framework that leverage AI to enhance the DSE efficiency and efficacy for GPUs. LUMINA extracts architectural knowledge from simulator code and performs sensitivity studies to automatically compose DSE rules,which are auto-corrected during exploration. A core component of LUMINA is a DSE Benchmark that comprehensively evaluates and enhances LLMs' capabilities across three fundamental skills required for architecture optimization, which provides a principled and reproducible basis for model selection and ensuring consistent architectural reasoning. In the design space with 4.7 million possible samples, LUMINA identifies 6 designs of better performance and area than an A100 GPU efficiently, using only 20 steps via LLM-assisted bottleneck analysis. In comparison, LUMINA achieves 17.5x higher than design space exploration efficiency, and 32.9% better designs (i.e. Pareto Hypervolume) than Machine-Learning baselines, showcasing its ability to deliver high-quality design guidance with minimal search cost.

cs.AR↗

arXiv2026

STEP: Detecting Audio Backdoor Attacks via Stability-based Trigger Exposure Profiling

With the widespread deployment of deep-learning-based speech models in security-critical applications, backdoor attacks have emerged as a serious threat: an adversary who poisons a small fraction of training data can implant a hidden trigger that controls the model's output while preserving normal behavior on clean inputs. Existing inference-time defenses are not well suited to the audio domain, as they either rely on trigger over-robustness assumptions that fail on transformation-based and semantic triggers, or depend on properties specific to image or text modalities. In this paper, we propose STEP (Stability-based Trigger Exposure Profiling), a black-box, retraining-free backdoor detector that operates under hard-label-only access. Its core idea is to exploit a characteristic dual anomaly of backdoor triggers: anomalous label stability under semantic-breaking perturbations, and anomalous label fragility under semantic-preserving perturbations. STEP profiles each test sample with two complementary perturbation branches that target these two properties respectively, scores the resulting stability features with one-class anomaly detectors trained on benign references, and fuses the two scores via unsupervised weighting. Extensive experiments across seven backdoor attacks show that STEP achieves an average AUROC of 97.92% and EER of 4.54%, substantially outperforming state-of-the-art baselines, and generalizes across model architectures, speech tasks, an open-set verification scenario, and over-the-air physical-world settings.

cs.CR↗

arXiv2026

The stringy geometry of integral cohomology in mirror symmetry

We examine the physical significance of torsion co-cycles in the cohomology of a projective Calabi-Yau three-fold for the (2,2) superconformal field theory (SCFT) associated to the non-linear sigma model with such a manifold as a target space. There are two independent torsion subgroups in the cohomology. While one is associated to an orbifold construction of the SCFT, the other encodes the possibility of turning on a topologically non-trivial flat gerbe for the NS-NS B-field. Inclusion of these data enriches mirror symmetry by providing a refinement of the familiar structures and points to a generalization of the duality symmetry, where the topology of the flat gerbe enters on the same footing as the topology of the underlying manifold.

hep-th↗

arXiv2026

Observation of Iso-Symmetric Structural and Lifshitz Transitions in Quasi-one-dimensional CrNbSe$_5$

Chalcogenides-rich transition metal compounds host a rich landscape of emergent quantum phenomena that are intimately governed by their quasi-one-dimensional chemical-bonding frameworks and their response to external perturbations such as pressure. Here, we report a pressure-induced iso-symmetric structural transition in the quasi-one-dimensional compound CrNbSe$_5$, in which the electronic ground state is controlled not by symmetry breaking but by a continuous reorganization of local bonding interactions. Applied pressure reversibly tunes CrNbSe$_5$ between semiconducting and semimetallic states, enabling access to low- and high-carrier electronic regimes through direct modulation of metal-chalcogen bonding. High-pressure single-crystal X-ray diffraction directly resolves the evolution of Cr-Se and Nb-Se bond distances, coordination polyhedra, and connectivity, revealing a fully reversible semimetal-semiconductor-semimetal transition driven by gradual yet cooperative bond rearrangements within a preserved crystallographic symmetry. In contrast to chemical substitution, which irreversibly alters composition and introduces disorder, pressure acts as a clean, continuous control parameter that reshapes the bonding landscape without disrupting structural symmetry. These results establish CrNbSe$_5$ as a model system for electronically driven phase switching via tunable chemical bonding, highlighting iso-symmetric bond reorganization as a powerful design principle for pressure-controlled electronic and spintronic functionalities.

cond-mat.mtrl-sci↗

arXiv2026

TrasMuon: Trust-Region Adaptive Scaling for Orthogonalized Momentum Optimizers

Muon-style optimizers leverage Newton-Schulz (NS) iterations to orthogonalize updates, yielding update geometries that often outperform Adam-series methods. However, this orthogonalization discards magnitude information, rendering training sensitive to step-size hyperparameters and vulnerable to high-energy bursts. To mitigate this, we introduce TrasMuon (\textbf{T}rust \textbf{R}egion \textbf{A}daptive \textbf{S}caling \textbf{Muon}). TrasMuon preserves the near-isometric geometry of Muon while stabilizing magnitudes through (i) global RMS calibration and (ii) energy-based trust-region clipping. We demonstrate that while reintroducing adaptive scaling improves optimization efficiency, it typically exacerbates instability due to high-energy outliers. TrasMuon addresses this by defining a trust region based on relative energy ratios, confining updates to a stable zone. Empirical experiments on vision and language models demonstrate that TrasMuon converges faster than baselines. Furthermore, experiments without warmup stages confirm TrasMuon's superior stability and robustness.

cs.LG↗

arXiv2026

RAC: Relation-Aware Cache Replacement for Large Language Models

The scaling of Large Language Model (LLM) services faces significant cost and latency challenges, making effective caching under tight capacity crucial. Existing cache replacement policies, from heuristics to learning-based methods, predominantly rely on limited-window statistics such as recency and frequency. We show these signals are not robust for real-world LLM workloads, which exhibit long reuse distances and sparse local recurrence. To address these limitations, we propose Relation-Aware Cache (RAC), an online eviction strategy that leverages semantic relations among requests to guide eviction decisions. RAC synthesizes two relation-aware signals: (1) Topical Prevalence, which aggregates access evidence at the topic level to capture long-horizon reuse; and (2) Structural Importance, which leverages local intra-topic dependency structure to discriminate entries by their future reuse value. Extensive evaluations show that RAC maintains high effectiveness across diverse workloads, consistently surpassing state-of-the-art baselines by 20%--30% in cache hit ratio.

cs.DB↗

arXiv2026

Automated Proof Generation for Rust Code via Self-Evolution

Ensuring correctness is crucial for code generation. Formal verification offers a definitive assurance of correctness, but demands substantial human effort in proof construction and hence raises a pressing need for automation. The primary obstacle lies in the severe lack of data-there is much fewer proofs than code snippets for Large Language Models (LLMs) to train upon. In this paper, we introduce SAFE, a framework that overcomes the lack of human-written proofs to enable automated proof generation of Rust code. SAFE establishes a self-evolving cycle where data synthesis and fine-tuning collaborate to enhance the model capability, leveraging the definitive power of a symbolic verifier in telling correct proofs from incorrect ones. SAFE also re-purposes the large number of synthesized incorrect proofs to train the self-debugging capability of the fine-tuned models, empowering them to fix incorrect proofs based on the verifier's feedback. SAFE demonstrates superior efficiency and precision compared to GPT-4o. Through tens of thousands of synthesized proofs and the self-debugging mechanism, we improve the capability of open-source models, initially unacquainted with formal verification, to automatically write proofs for Rust code. This advancement leads to a significant improvement in performance, achieving a 52.52% accuracy rate in a benchmark crafted by human experts, a significant leap over GPT-4o's performance of 14.39%.

cs.SE↗

arXiv2026

Improving Data and Reward Design for Scientific Reasoning in Large Language Models

Solving open-ended science questions remains challenging for large language models, particularly due to inherently unreliable supervision and evaluation. The bottleneck lies in the data construction and reward design for scientific post-training. We develop a large-scale, systematic data processing pipeline that transforms heterogeneous open-source science data into Dr. SCI dataset, which comprises of 1M questions across eight STEM subjects, with explicit verifiable/open-ended splits, scalable difficulty annotation, and fine-grained rubrics that operationalize evaluation for open-ended answers. Building on this dataset, we propose the Dr. SCI post-training pipeline, which redesigns the standard SFT -> RL workflow through three components: (i) Exploration-Expanding SFT, which broadens the model's reasoning pattern coverage prior to RL; (ii) Dynamic Difficulty Curriculum, which adapts training data to the model's evolving scientific capability; and (iii) SciRubric-Guided RL, which enables stable reinforcement learning on open-ended scientific questions via rubric-based evaluation with explicit answer correctness. Qwen3-4B-Base trained using Dr. SCI pipeline achieves 63.2 on GPQA-diamond and 32.4 on GPQA-general, consistently improves over strong post-trained baselines such as o1-mini and GPT-4o, demonstrating substantial gains in scientific reasoning, especially in open-ended settings.

cs.CL↗

arXiv2026

LocDreamer: World Model-Based Learning for Joint Indoor Tracking and Anchor Scheduling

Accurate, resource-efficient localization and tracking enables numerous location-aware services in next-generation wireless networks. However, existing machine learning-based methods often require large labeled datasets while overlooking spectrum and energy efficiencies. To fill this gap, we propose LocDreamer, a world model (WM)-based framework for joint target tracking and scheduling of localization anchors. LocDreamer learns a WM that captures the latent representation of the target motion and localization environment, thereby generating synthetic measurements to imagine arbitrary anchor deployments. These measurements enable imagination-driven training of both the tracking model and the reinforcement learning (RL)-based anchor scheduler that activates only the most informative anchors, which significantly reduce energy and signaling costs while preserving high tracking accuracy. Experiments on a real-world indoor dataset demonstrate that LocDreamer substantially improves data efficiency and generalization, outperforming conventional Bayesian filter with random scheduling by 37% in tracking accuracy, and achieving 86% of the accuracy of same model trained directly on real data.

eess.SP↗

arXiv2026

Synthesizing File-Level Data for Unit Test Generation with Chain-of-Thoughts via Self-Debugging

Automatic unit test (UT) generation is essential for software quality assurance, but existing approaches--including symbolic execution, search-based approaches, and recent LLM-based generators--struggle to produce human-quality tests with correct, meaningful assertions and reliable chain-of-thought (CoT) explanations. We identify a gap in UT training data: repository-mined tests lack developer CoTs, while LLM-distilled CoTs are often incorrect or incomplete. To address this issue, we propose a novel data-distillation approach that uses self-debugging to produce high-quality UT training examples paired with faithful CoTs. Our approach combines (1) guided test repair, a heuristic loop (error-, failure-, and coverage-focused steps) that asks the used model to diagnose and iteratively fix generated tests, and (2) CoT compression, which compacts original and debugging CoTs into concise explanations that directly justify correct tests. We apply this pipeline to a large corpus of open-source projects to construct a dataset of 74,518 high-quality examples, and then use it for supervised fine-tuning of a base model. An empirical evaluation shows that the fine-tuned model achieves high UT generation effectiveness: it attains a pass rate of 36.17% on test assertions, a branch coverage of 43.90%, and a mutation score of 88.66%, substantially higher than state-of-the-art commercial models like o4-mini.

cs.SE↗

arXiv2026

Divide and Conquer: Multimodal Video Deepfake Detection via Cross-Modal Fusion and Localization

This paper presents a system for detecting fake audio-visual content (i.e., video deepfake), developed for Track 2 of the DDL Challenge. The proposed system employs a two-stage framework, comprising unimodal detection and multimodal score fusion. Specifically, it incorporates an audio deepfake detection module and an audio localization module to analyze and pinpoint manipulated segments in the audio stream. In parallel, an image-based deepfake detection and localization module is employed to process the visual modality. To effectively leverage complementary information across different modalities, we further propose a multimodal score fusion strategy that integrates the outputs from both audio and visual modules. Guided by a detailed analysis of the training and evaluation dataset, we explore and evaluate several score calculation and fusion strategies to improve system robustness. Overall, the final fusion-based system achieves an AUC of 0.87, an AP of 0.55, and an AR of 0.23 on the challenge test set, resulting in a final score of 0.5528.

cs.MM↗

arXiv2026

Synthesis of Monolayer Ice on a Hydrophobic Metal Surface

Understanding water-metal interactions is central to disciplines spanning catalysis, electrochemistry, and atmospheric science. Monolayer ice phases are well established on hydrophilic surfaces, where strong water-substrate interactions stabilize ordered hydrogen-bond networks. In contrast, their formation on hydrophobic metals has been deemed ther-modynamically unfavourable, with water typically assembling into amorphous films, three-dimensional crystallites, or interlocked bilayer ice. Here, we demonstrate the synthesis of a monolayer ice phase on the hydrophobic Au(111) surface using a low-energy-electron-assisted growth method. Combined experimental characterizations including low-energy electron diffraction, angle-resolved photoemission spectroscopy, and X-ray photoelectron spectroscopy, complemented by first-principles calculations, prove that the monolayer ice phase composes of intact water molecules. This approach provides a generalizable strategy for stabilizing ordered two-dimensional ice on inert substrates and offers new insight into the interplay between water and low-energy electrons at hydrophobic interfaces.

cond-mat.mtrl-sci↗

arXiv2026

Bridging the Arithmetic Gap: The Cognitive Complexity Benchmark and Financial-PoT for Robust Financial Reasoning

While Large Language Models excel at semantic tasks, they face a critical bottleneck in financial quantitative reasoning, frequently suffering from "Arithmetic Hallucinations" and a systemic failure mode we term "Cognitive Collapse". To strictly quantify this phenomenon, we introduce the Cognitive Complexity Benchmark (CCB), a robust evaluation framework grounded in a dataset constructed from 95 real-world Chinese A-share annual reports. Unlike traditional datasets, the CCB stratifies financial queries into a three-dimensional taxonomy, Data Source, Mapping Difficulty, and Result Unit, enabling the precise diagnosis of reasoning degradation in high-cognitive-load scenarios. To address these failures, we propose the Iterative Dual-Phase Financial-PoT framework. This neuro-symbolic architecture enforces a strict architectural decoupling: it first isolates semantic variable extraction and logic formulation, then offloads computation to an iterative, self-correcting Python sandbox to ensure deterministic execution. Evaluation on the CCB demonstrates that while standard Chain-of-Thought falters on complex tasks, our approach offers superior robustness, elevating the Qwen3-235B model's average accuracy from 59.7\% to 67.3\% and achieving gains of up to 10-fold in high-complexity reasoning tasks. These findings suggest that architectural decoupling is a critical enabling factor for improving reliability in financial reasoning tasks, providing a transferable architectural insight for precision-critical domains that require tight alignment between semantic understanding and quantitative computation.

cs.AI↗

arXiv2026

MSCCL++: Rethinking GPU Communication Abstractions for AI Inference

AI applications increasingly run on fast-evolving, heterogeneous hardware to maximize performance, but general-purpose libraries lag in supporting these features. Performance-minded programmers often build custom communication stacks that are fast but error-prone and non-portable. This paper introduces MSCCL++, a design methodology for developing high-performance, portable communication kernels. It provides (1) a low-level, performance-preserving primitive interface that exposes minimal hardware abstractions while hiding the complexities of synchronization and consistency, (2) a higher-level DSL for application developers to implement workload-specific communication algorithms, and (3) a library of efficient algorithms implementing the standard collective API, enabling adoption by users with minimal expertise. Compared to state-of-the-art baselines, MSCCL++ achieves geomean speedups of $1.7\times$ (up to $5.4\times$) for collective communication and $1.2\times$ (up to $1.38\times$) for AI inference workloads. MSCCL++ is in production of multiple AI services provided by Microsoft Azure, and has also been adopted by RCCL, the GPU collective communication library maintained by AMD. MSCCL++ is open source and available at https://github.com/microsoft/mscclpp . Our two years of experience with MSCCL++ suggests that its abstractions are robust, enabling support for new hardware features, such as multimem, within weeks of development.

cs.DC↗

arXiv2026

Attack-Resistant Watermarking for AIGC Image Forensics via Diffusion-based Semantic Deflection

Protecting the copyright of user-generated AI images is an emerging challenge as AIGC becomes pervasive in creative workflows. Existing watermarking methods (1) remain vulnerable to real-world adversarial threats, often forced to trade off between defenses against spoofing and removal attacks; and (2) cannot support semantic-level tamper localization. We introduce PAI, a training-free inherent watermarking framework for AIGC copyright protection, plug-and-play with diffusion-based AIGC services. PAI simultaneously provides three key functionalities: robust ownership verification, attack detection, and semantic-level tampering localization. Unlike existing inherent watermark methods that only embed watermarks at noise initialization of diffusion models, we design a novel key-conditioned deflection mechanism that subtly steers the denoising trajectory according to the user key. Such trajectory-level coupling further strengthens the semantic entanglement of identity and content, thereby further enhancing robustness against real-world threats. Moreover, we also provide a theoretical analysis proving that only the valid key can pass verification. Experiments across 12 attack methods show that PAI achieves 98.43\% verification accuracy, improving over SOTA methods by 37.25\% on average, and retains strong tampering localization performance even against advanced AIGC edits. Our code is available at https://github.com/QingyuLiu/PAI.

cs.CR↗