arXiv subjects

Teng Ma

Publications and source records attributed to Teng Ma.

At least 19 recordsLinked to original sources

arXiv2026

Static-Recoil Factorization in Heavy-Baryon Chiral EFTs

Extending heavy-baryon $\chi$PT to higher-spin resonances introduces unphysical lower-spin admixtures, leading to costly path-integral projections. To resolve this, we develop an on-shell implementation of the heavy-baryon expansion based on recoil-channel partial-wave eigenoperators. These eigenoperators separate the static mass dependence from recoil structures and assign chiral order directly in amplitude space. The resulting static--recoil factorization systematically generates non-redundant local operators for heavy-baryon sectors involving higher-spin resonances. Flavor structures and identical-particle constraints are imposed through a new linear-algebraic reduction that uses von Neumann alternating projections to extract the corresponding common physical subspace. Beyond this specific application, the framework can be broadly applied to general nonrelativistic effective field theories.

hep-ph↗

arXiv2026

Local Minimum of Spin-Sector Magic at the CP-Conserving Point in Low-Energy Neutron-Proton Scattering

We study Magic generation in elastic neutron-proton scattering within a leading low-energy spin-sector ansatz that retains the one-pion-exchange spin structures and treats each scattering direction as a conditional two-qubit spin map. We show that the direction-averaged Magic is locally minimized at the CP-conserving (CPC) point $\bar\theta=0$ at the Clifford point $f_{\rm CPC}=\pi/4$, and for the representative non-Clifford CPC backgrounds analyzed here. At $f_{\rm CPC}=\pi/4$, the CPC spin map reduces to SWAP up to a phase and therefore generates zero Magic from stabilizer inputs. We further evaluate the complete spin-sector Magic functional by averaging over all 60 two-qubit stabilizer inputs and over scattering directions, and find that the curvature at $\bar\theta=0$ is positive only within specific windows of the effective CPC phase $f_{\rm CPC}$. These results identify the CPC point as a local Magic minimum within the restricted low-energy spin sector considered here.

hep-ph↗

arXiv2026

SAC: Disaggregated KV Cache System for Sparse Attention LLMs with CXL

The scaling of LLMs toward long-context inference has shifted the primary serving system bottleneck from computation to memory capacity. Traditional solutions for dense attention models rely on RDMA-based disaggregated memory pools, which perform coarse-grained fetching of the entire prefix KV cache from remote storage to local memory before decoding. However, this approach is fundamentally inefficient for emerging sparse attention models. While only a small fraction of KV entries are active during decoding, these systems still fetch the full KV cache locally, leading to severe transmission bottlenecks and local memory wastage. To address this, we propose SAC, the first efficient disaggregated KV cache system optimized for sparse attention models. By leveraging the low-latency, cache-line granularity load/store semantics of Compute Express Link (CXL), SAC fetches only the required top-k KV entries on demand during inference. Evaluations on DeepSeek-V3.2 using SGLang show that SAC achieves 2.1x higher throughput, 9.7x lower TTFT, and 1.8x lower TBT compared to RDMA-based baselines, establishing CXL-based disaggregation as the superior infrastructure for emerging sparse attention models.

cs.DC↗

arXiv2026

Towards Unified Song Generation and Singing Voice Conversion with Accompaniment Co-Generation

While song generation and singing voice conversion (SVC) have evolved significantly, they have long been developed isolated: the former lacks zero-shot speaker cloning, while the latter overlooks vocal-accompaniment synergy. To bridge this gap, we propose UniSinger, the first end-to-end framework unifying speaker cloning song generation and accompaniment co-generation SVC. Building on the multimodal diffusion transformer, we construct a unified speaker embedding space transferring speaker representation from SVC to song generation, endowing fine-grained cross-task timbre control. To mitigate multi-task optimization conflicts, we design a curriculum learning strategy using task-specific modality masking to guide the model to gradually master the generative mechanisms among semantic content, vocal timbre, and accompaniment. Experiments show state-of-the-art performance on both tasks and realizes complementary benefits, offering new possibilities for intelligent music production.

cs.SD↗

arXiv2026

ShotCrop$^3$: Cropping Human-Centric Images into Cinematic Triple-Shot Compositions

Prior work on aesthetic composition typically produces a single aesthetically pleasing crop, overlooking the narrative value of composing multiple shots from one scene. In practice, multi-shot composition is critical for downstream creative workflows: commercial posters often require multiple crops with different emphases (e.g., context, subject, and emotion/product details) to present key story beats. Therefore, we propose \textbf{Triple-Shot Compositions (TSC)}, a composition task that generates a three-shot set -- establishing, medium, and close-up -- from a single human-centric image, each paired with a brief shot description to support visual narration. To learn TSC with limited expert annotations, we introduce \textbf{ShotCrop} which undergoes a three-stage training process: it first applies Chain-of-Thought supervised fine-tuning to establish basic reasoning and aesthetic shot-cropping skills, then performs semi-supervised fine-tuning with high-confidence pseudo labels to further enhance aesthetic capability, and is finally optimized with Group Relative Policy Optimization for \textbf{ShotCrop} (GRPO-S) using a composite reward tailored for it. Specifically, our pseudo-labeling strategy combines MLLM-based scoring, aesthetic assessment, and CLIP similarity to retain high-confidence training signals. In addition, we present TSC-Bench, a benchmark of 1.2k expert-annotated test cases. Notably, ShotCrop achieves an average improvement of \textbf{2.82} times over GPT-5 in shot localization accuracy.

cs.CV↗

arXiv2026

Heavy Axion from a Confining Mirror GUT

We propose a new framework for solving the strong CP problem via a heavy axion, using mirror symmetry and grand unification. The mirror GUT sector remains unbroken and dynamically generates a calculable heavy mass scale via confinement without fine tuning. Models in this class feature a heavy axion, whose potential is less sensitive to Planck scale corrections, as well as a rich hidden sector from the confined mirror GUT. The solution to the strong CP problem remains unspoiled by the presence of additional phases in the GUT Yukawas, yet allowing the possibility of electric dipole moments within the reach of future experiments. Our proposal offers new directions in GUT model building, axion phenomenology, dark matter and cosmology.

hep-ph↗

arXiv2026

DisagFusion: Asynchronous Pipeline Parallelism and Elastic Scheduling for Disaggregated Diffusion Serving

Diffusion-based generation is increasingly powering production content pipelines; however, deploying these models at scale remains a significant challenge. Model weights frequently exceed the memory capacity of commodity GPUs, while the encoder, diffusion transformer (DiT), and decoder stages exhibit highly imbalanced computational and memory footprints. A natural remedy is disaggregated serving-running stages as separate services on heterogeneous GPUs-yet this introduces new bottlenecks, including stage handoff overheads and fast-changing workloads that make cross-stage provisioning and scheduling brittle. This paper presents DisagFusion, enabling asynchronous pipeline parallelism and elastic scheduling for disaggregated diffusion serving. First, DisagFusion introduces asynchronous pipeline parallelism that overlaps computation and stage-to-stage communication to reduce pipeline bubbles and mitigate network jitter. Second, DisagFusion employs a hybrid instance scheduling strategy that combines lightweight performance prediction with runtime feedback to continuously rebalance instance ratio across stages under workload shifts. We implement DisagFusion and evaluate it with modern diffusion models. Compared to a monolithic baseline, DisagFusion improves throughput by 3.4x-20.5x and reduces end-to-end latency by 18.5x, while enabling flexible, cost-efficient deployment across heterogeneous GPUs.

cs.DC↗

arXiv2026

Entanglement Maximization and Symmetry Selection in Composite Higgs Models

Recent developments suggest that the extremization of quantum entanglement may provide a useful organizing principle for strong dynamics. While entanglement suppression characterizes low-energy QCD, we investigate the role of entanglement maximization in the electroweak symmetry breaking sector. Focusing on the Composite Higgs Model, we analyze the process $hh \to t\bar{t}$ by treating the fermionic helicity space as a bipartite quantum system. Maximal entanglement imposes nontrivial constraints on the fermionic effective theory and leads to two simple symmetry structures in the top sector. One is the Maximal Symmetry branch, characterized by the vanishing of the Higgs-dependent form factor $\Pi_1$ and the finiteness of the Higgs potential. The other is a generalized $Z_2$-matching branch relating the left- and right-handed top sectors. Our results establish a quantitative connection between entanglement structure and the naturalness of electroweak symmetry breaking, and suggest that the symmetry patterns of the strong sector may be understood from the perspective of entanglement extremization.

hep-ph↗

arXiv2026

Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference

Mixture-of-Experts (MoE) serving relies on wide expert parallelism (EP) to aggregate the memory capacity and bandwidth of many GPUs within one inference instance. This efficiency comes with a systems cost: every decoding step depends on token dispatch and combination across all active EP ranks, so even one rank failure can disrupt the entire service. Existing EP stacks handle such failures poorly because they treat membership as a fixed configuration established at initialization. The same rank set determines communicator state, expert placement, and the routing metadata baked into CUDA execution graphs, leaving the system with no way to shrink around a failure while keeping the instance valid. This paper argues that partial-failure tolerance should instead be formulated as a live EP validity problem. We present EEP, a communication and runtime substrate that represents membership as explicit, mutable runtime state. EEP repairs the specific state invalidated by a fault: it restores peer reachability without rebuilding the communication substrate, repairs lost expert coverage through a bandwidth-aware hierarchy, and reintegrates repaired ranks without forcing healthy ranks to recapture their CUDA graphs. We implement EEP in an EP serving stack integrated with SGLang and evaluate it under steady-state serving, failure recovery, and rank reintegration. The results show that explicit mutable membership preserves the steady-state fast path, staying within 4.4% of a fixed-membership DeepEP baseline under static serving, while turning a local rank fault from whole-instance downtime into two bounded interruptions. On a single-rank failure workload, EEP incurs an 11s recovery pause and an 8s reintegration pause, and restores throughput to within 95% of the pre-fault level within 52s, whereas a fixed-membership full-restart baseline remains unavailable until 348s.

cs.DC↗

arXiv2026

ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL

Agentic reinforcement learning (RL) is reshaping LLM post-training, but end-to-end training time is dominated by compute-intensive, multi-turn rollouts whose resource demand varies significantly across training steps. Resource-fixed systems cannot adapt to this variation, while resource-elastic approaches that provision external GPUs on demand suffer from high allocation overhead and limited availability. We observe that serving clusters leave substantial GPU compute and memory idle, and propose cooperative elasticity: sharing already-deployed serving GPUs with rollout workloads to provide on-demand elastic capacity. Realizing this is non-trivial, as it must preserve serving SLOs under bursty traffic while minimizing cross-cluster communication overhead. We present ROSE, a system that realizes cooperative elasticity for agentic RL post-training, comprising three components: (1) an SLO-safe co-serving executor that co-locates heterogeneous serving and rollout models on the same GPUs, dynamically sharing memory and compute while preserving serving SLOs; (2) a cross-cluster weight transfer engine that leverages shard-aware routing and weight sparsity for fast synchronization; and (3) an elastic rollout scheduler that dynamically routes rollouts across dedicated and opportunistic serving GPUs. Experiments across multiple model sizes and cluster scales show that ROSE improves end-to-end throughput by 1.3 - 3.3 x over resource-fixed baselines and reduces rollout time by 1.2 - 1.5 x over resource-elastic baselines, with no serving SLO violations.

cs.DC↗

arXiv2026

SiriusHelper: An LLM Agent-Based Operations Assistant for Big Data Platforms

Big data platforms are widely used in modern enterprises, and an in-production intelligent assistant is increasingly important to help users quickly find actionable guidance and reduce operational burden. While recent LLM+RAG assistants provide a natural interface, they face practical challenges in real deployments: limited scenario coverage across both general consultation and domain-specific troubleshooting workflows, inefficient knowledge access due to inadequate multi-hop retrieval and flat knowledge organization, and high maintenance cost because escalated tickets are unstructured and hard to convert into assistant improvements and reusable SOPs. In this paper, we present SiriusHelper, a deployed intelligent assistant for big data platforms. SiriusHelper serves as a unified online assistant that automatically identifies user intent and routes queries to the right handling path, including dedicated expert workflows for specialized scenarios (e.g., SQL execution diagnosis). To support complex troubleshooting, SiriusHelper combines a DeepSearch-driven mechanism with a priority-based hierarchical knowledge base to enable multi-hop retrieval without context overload, thus improving answer reliability and latency. To reduce expert overhead, SiriusHelper further introduces automated ticket understanding and SOP distillation: it diagnoses the assistant failure reason (e.g., missing knowledge or wrong routing) and extracts domain-specific SOPs to continuously enrich the knowledge base. Experiments and online deployment on Tencent Big Data platform show that SiriusHelper outperforms representative alternatives and reduces online ticket volume by 20.8\%.

cs.DB↗

arXiv2026

UniSonate: A Unified Model for Speech, Music, and Sound Effect Generation with Text Instructions

Generative audio modeling has largely been fragmented into specialized tasks, text-to-speech (TTS), text-to-music (TTM), and text-to-audio (TTA), each operating under heterogeneous control paradigms. Unifying these modalities remains a fundamental challenge due to the intrinsic dissonance between structured semantic representations (speech/music) and unstructured acoustic textures (sound effects). In this paper, we introduce UniSonate, a unified flow-matching framework capable of synthesizing speech, music, and sound effects through a standardized, reference-free natural language instruction interface. To reconcile structural disparities, we propose a novel dynamic token injection mechanism that projects unstructured environmental sounds into a structured temporal latent space, enabling precise duration control within a phoneme-driven Multimodal Diffusion Transformer (MM-DiT). Coupled with a multi-stage curriculum learning strategy, this approach effectively mitigates cross-modal optimization conflicts. Extensive experiments demonstrate that UniSonate achieves state-of-the-art performance in instruction-based TTS (WER 1.47%) and TTM (SongEval Coherence 3.18), while maintaining competitive fidelity in TTA. Crucially, we observe positive transfer, where joint training on diverse audio data significantly enhances structural coherence and prosodic expressiveness compared to single-task baselines. Audio samples are available at https://qiangchunyu.github.io/UniSonate/.

eess.AS↗

arXiv2026

One-shot learning for the complex dynamical behaviors of weakly nonlinear forced oscillators

Extrapolative prediction of complex nonlinear dynamics remains a central challenge in engineering. This study proposes a one-shot learning method to identify global frequency-response curves from a single excitation time history by learning governing equations. We introduce MEv-SINDy (Multi-frequency Evolutionary Sparse Identification of Nonlinear Dynamics) to infer the governing equations of non-autonomous and multi-frequency systems. The methodology leverages the Generalized Harmonic Balance (GHB) method to decompose complex forced responses into a set of slow-varying evolution equations. We validated the capabilities of MEv-SINDy on two critical Micro-Electro-Mechanical Systems (MEMS). These applications include a nonlinear beam resonator and a MEMS micromirror. Our results show that the model trained on a single point accurately predicts softening/hardening effects and jump phenomena across a wide range of excitation levels. This approach significantly reduces the data acquisition burden for the characterization and design of nonlinear microsystems.

cs.LG↗

arXiv2026

A Generalized Method for Spatial Operations on Physical Properties of Matter

The physical properties of matter are typically described by coefficient matrices governed by crystal symmetry. Applying spatial operations, such as rotation, inversion, and mirror, to these matrices provides an effective approach for investigating material properties. However, the diversity of coefficient matrix types complicates their transformation via simple matrix multiplication, and existing methods suffer from cumbersome notation, high computational cost, and lack of intuitive interpretation. Moreover, as coefficient matrices grow in size, conventional approaches become increasingly inadequate. We present a generalized ``input-coefficient-output (ICO)" approach for constructing spatial operation matrices applicable to coefficient matrices across diverse physical systems, including but not limited to high-order nonlinear optics, elastic mechanics, electricity and magnetism. Our approach offers a concise formalism that enables intuitive reasoning about spatial transformations while delegating intensive computations to computational tools, which is analogous to the role of Feynman diagrams in facilitating understanding in physics. This method also offers valuable insights for future theoretical and experimental research.

cond-mat.mtrl-sci↗

arXiv2026

TENT: A Declarative Slice Spraying Engine for Performant and Resilient Data Movement in Disaggregated LLM Serving

Modern GPU clusters rely on complex, heterogeneous interconnects. As large language model (LLM) serving shifts toward agentic reasoning, KVCache becomes a first-class mobile asset, driving frequent migrations and massive elephant flows that dominate the execution critical path. Operating Mooncake Transfer Engine (TE) on thousands of GPUs exposed a fundamental flaw in existing frameworks: imperative, early-binding path selection. This rigidity results in state-blind striping that ignores congestion and grey failures, leading to bandwidth stranding. This also produces operational fragility where routine faults require manual intervention. We present TENT, a declarative orchestration engine that decouples transfer intent from physical execution. By abstracting interconnects into a unified resource pool, TENT shifts path resolution from initialization to slice-time late binding. Applications simply declare transfer intents, while TENT dynamically "sprays" fine-grained slices across rails based on real-time telemetry and predictive cost modeling. This orchestration eliminates head-of-line (HoL) blocking and enables transparent, sub-50 ms self-healing by rerouting slices around failures or degradations without application-level intervention. TENT serves as the production data plane for LLM inference and reinforcement learning (RL) pipelines at multiple industrial clusters. Our evaluation shows that TE+ outperforms state-of-the-art baselines, including Mooncake TE, NIXL, and UCCL. In LLM inference with SGLang HiCache, TE+ achieves up to 1.36x higher throughput and 26% lower P90 time-to-first-token (TTFT) than Mooncake TE. In RL pipelines, TENT accelerates parameter updates in Moonshot Checkpoint Engine by 20-26%.

cs.DC↗

arXiv2026

V-CAST: Video Curvature-Aware Spatio-Temporal Pruning for Efficient Video Large Language Models

Video large language models (VideoLLMs) show strong capability in video understanding, yet long-context inference is still dominated by massive redundant visual tokens in the prefill stage. We revisit token compression for VideoLLMs under a tight budget and identify a key bottleneck, namely insufficient spatio-temporal information coverage. Existing methods often introduce discontinuous coverage through coarse per-frame allocation or scene segmentation, and token merging can further misalign spatio-temporal coordinates under MRoPE-style discrete (t,h,w) bindings. To address these issues, we propose V-CAST (Video Curvature-Aware Spatio-Temporal Pruning), a training-free, plug-and-play pruning policy for long-context video inference. V-CAST casts token compression as a trajectory approximation problem and introduces a curvature-guided temporal allocation module that routes per-frame token budgets to semantic turns and event boundaries. It further adopts a dual-anchor spatial selection mechanism that preserves high-entropy visual evidence without attention intervention, while keeping retained tokens at their original coordinates to maintain positional alignment. Extensive experiments across multiple VideoLLMs of different architectures and scales demonstrate that V-CAST achieves 98.6% of the original performance, outperforms the second-best method by +1.1% on average, and reduces peak memory and total latency to 86.7% and 86.4% of vanilla Qwen3-VL-8B-Instruct.

cs.CV↗

arXiv2026

Pooling Engram Conditional Memory in Large Language Models using CXL

Engram conditional memory has emerged as a promising component for LLMs by decoupling static knowledge lookup from dynamic computation. Since Engram exhibits sparse access patterns and supports prefetching, its massive embedding tables are well-suited for offloading to lower-tier memory. In this paper, we propose using Compute Express Link (CXL) memory pool for Engram storage. Compared to RDMA, CXL provides fine-grained and low-latency access required by minimal and discrete retrieval patterns of Engram. We integrate the CXL-based Engram pool into SGLang, achieving near-DRAM end-to-end performance. This provides a scalable and cost-efficient storage solution for future Engram-integrated LLMs without compromising inference performance.

cs.AR↗

arXiv2026

Adaptive Multi-Objective Tiered Storage Configuration for KV Cache in LLM Service

The memory-for-computation paradigm of KV caching is essential for accelerating large language model (LLM) inference service, but limited GPU high-bandwidth memory (HBM) capacity motivates offloading the KV cache to cheaper external storage tiers. While this expands capacity, it introduces the challenge of dynamically managing heterogeneous storage resources to balance cost, throughput, and latency under varying workloads. We formulate this as a multi-objective optimization problem: identifying the Pareto frontier across these metrics within the storage configuration space. Using a high-fidelity end-to-end simulator, we observe that the objective functions are non-analytic and exhibit complex variable coupling, making the Pareto frontier difficult to approximate analytically. To obtain the frontier, we introduce Kareto, a KV-cache Adaptive REsource managemenT Optimizer. Kareto leverages a diminishing-return-guided pruning method to efficiently navigate the large configuration space and approximate the Pareto frontier. Additionally, it incorporates a fine-grained adaptive tuner that uses eviction policies in tier storage and KV block access patterns for group-specific cache management, improving cache efficiency. Experiments on real-world traces show that Kareto adapts to workload and can identify configurations of better cost efficiency, covering static strategies. Compared to the fixed setup with 1024 GB DRAM, Kareto can improve throughput by up to 9.3%, or reduce latency by up to 58.3%, or lower cost by up to 20.2% under respective optimization objectives.

cs.AR↗