arXiv subjects

Chao Deng

Publications and source records attributed to Chao Deng.

At least 19 recordsLinked to original sources

arXiv2026

KnowGPO: Knowledge-Grouped Policy Optimization for Search Agents

Deep search agents solve knowledge-intensive tasks through long-horizon, multi-turn interactions, where assigning credit to individual search decisions is fundamentally difficult. Outcome-based reinforcement learning provides only a single terminal reward shared across an entire trajectory. Recent process-supervision methods aim to assign credit at finer granularity, but they typically anchor credit to either surface-level trajectory states or model-internal signals, both of which become unreliable in deep search where observations are heterogeneous and retrieved contexts are long and noisy. We propose KnowGPO, a reinforcement learning framework that aligns trajectories by the task-relevant knowledge they have acquired rather than by their surface observations. For each query, KnowGPO decomposes the information required to derive the answer into atomic knowledge conditions, and represents the state of a trajectory at each step by the subset of conditions it has satisfied. Trajectories sharing the same knowledge state are grouped for within-group comparison, and the incremental knowledge gain between consecutive steps serves as an environment-grounded process reward. Experiments on Xbench, BrowseComp-ZH, and GAIA show KnowGPO consistently outperforms outcome-only (GRPO) and process-supervision baselines (GiGPO, IGPO) by 3.8 to 4.2 points, with our 8B model rivaling or exceeding most 32B open-source systems.

cs.AI↗

arXiv2026

DataEvolver: Automatic Data Preparation for Large Language Models through Multi-Level Self-Evolving

High-quality training data is essential to large language models (LLMs) and typically requires extensive and costly manual curation. Automatic data preparation systems support reusable operators and task-driven workflows, yet adapting these workflows to the desired data quality remains challenging. High-quality examples provide a practical specification for guiding this adaptation. We introduce DataEvolver, a data preparation system that automatically constructs pipelines through multi-level self-evolution. At the operator level, it expands the operator library and refines logical plans to resolve dependency conflicts. At the pipeline level, it instantiates these plans into code and uses trial-output feedback to refine the pipeline toward seed-specified quality requirements. Experiments on seven benchmarks show that DataEvolver substantially improves data quality and achieves an average gain of approximately 10 score points in downstream LLM performance compared with training on original data, highlighting new opportunities for the iterative co-evolution of LLMs and data.

cs.DB↗

arXiv2026

Vesta: Queueing-Aware LLM Scheduling for Real-World Workloads

Agentic applications generate multi-turn sequences of LLM requests, with each new turn extending the context accumulated from prior interactions and tool calls. Prefix caching reuses the KV cache of earlier turns and avoids redundant prefill, but routing every turn to its cache-affine instance can create severe local queues while capacity remains idle elsewhere. Remote KV cache access relaxes this tension, yet migrating execution adds transfer work and is harmful when the cluster is already congested. Thus, deciding where to execute a request requires reasoning jointly about cache locality, queueing, and the current operating point of the cluster. We present Vesta, a queueing-aware scheduler for distributed agentic LLM serving. Vesta assigns each session a stable holder for its authoritative KV cache while independently selecting an executor for every request. Grounded in classical queueing theory, its objective accounts for backlog, reusable prefixes, KV cache transfer, and the queueing externality of additional service demand; it migrates execution to recover idle capacity when beneficial and favors placements that add the least prefill and transfer work as congestion rises. Vesta also migrates reusable prefixes across length buckets to avoid cold starts. On real-world agentic workloads, Vesta improves throughput by 14.0\% and reduces mean TTFT by 57.8\% over the state-of-the-art scheduler in a PD-colocated deployment. In PD-disaggregated deployments, it reduces mean TTFT by up to 18.1\% and P99 TTFT by up to 29.0\% at matched throughput.

cs.DC↗

arXiv2026

Multi-Probe Zero Collision Hash (MPZCH): Mitigating Embedding Collisions and Enhancing Model Freshness in Large-Scale Recommenders

Embedding tables are critical components of large-scale recommendation systems, facilitating the efficient mapping of high-cardinality categorical features into dense vector representations. However, as the volume of unique IDs expands, traditional hash-based indexing methods suffer from collisions that degrade model performance and personalization quality. We present Multi-Probe Zero Collision Hash (MPZCH), a novel indexing mechanism based on linear probing that effectively mitigates embedding collisions. With reasonable table sizing, it often eliminates these collisions entirely while maintaining production-scale efficiency. MPZCH utilizes auxiliary tensors and high-performance CUDA kernels to implement configurable probing and active eviction policies. By retiring obsolete IDs and resetting reassigned slots, MPZCH prevents the stale embedding inheritance typical of hash-based methods, ensuring new features learn effectively from scratch. Despite its collision-mitigation overhead, the system maintains training QPS and inference latency comparable to existing methods. Rigorous online experiments demonstrate that MPZCH achieves zero collisions for user embeddings and significantly improves item embedding freshness and quality. The solution has been released within the open-source TorchRec library for the broader community.

cs.LG↗

arXiv2026

IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis

Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information. To address these issues, we propose IterSynth, a role-decoupled and summary-based paradigm that alternates between a Planner for identifying information needs and a Synthesizer for integrating evidence into an evolving summary state. This design separates planning from synthesis while using the summary as the persistent state of search, reducing both capability coupling and context noise. To train IterSynth effectively, we further introduce Role-Decoupled Policy Optimization (RDPO) for reinforcement learning, which combines terminal outcome rewards with turn-level rubric evaluations and computes role-specific advantages for more precise credit assignment. Experiments on five long-horizon deep-search benchmarks such as BrowseComp and Xbench-DS show that IterSynth-8B achieves an average score of 50.7, surpassing the strongest prior $\leq$8B agent by +4.2\%. Moreover, IterSynth serves as a model-agnostic prompting paradigm, delivering substantial zero-shot gains over ReAct and similar prompting paradigms on frontier proprietary models.

cs.CL↗

arXiv2026

Global large solutions of the Cauchy problem for the NS-NPP equations with Fujita-Kato type initial data

In this paper, we prove the global well-posedness of the Cauchy problem for the NS-NPP equations with some large Fujita-Kato type initial data. Specifically, we show that there exist two positive constants $c_{0}$ and $C_{0}$ such that if the initial data $(u_{0}, N_{0}, P_{0})$ satisfies the following condition: \begin{equation*} \left(\|u_0\|_{\dot{H}^{-1+\frac{d}{2}}}+\|N_{0}-P_{0}\|_{\dot{H}^{-2+\frac{d}{2}}}\right)\exp\left\{C_{0}\big(\|N_{0}+P_{0}\|_{\dot{H}^{-2+\frac{d}{2}}}^2+1\big)\right\} \leq c_{0}, \end{equation*} then the NS-NPP equations admits a unique global solution. This result implies global existence of solutions without any smallness conditions imposed on the sum of initial particle densities of negative and positive electric charge in the framework of Sobolev spaces.

math.AP↗

arXiv2026

Digital Self-Interference Cancellation in Full-Duplex Radios: A Fundamental Limit Perspective

D-SIC is of crucial importance for the implementation of IBFD radios. Unfortunately, the achievable performance limit remains underexplored. To fill this gap, in this paper we aim to explore the performance limit, i.e., the minimum residual self-interference (RSI) of the most commonly used PH canceller, and provide the achievable pilot design accordingly. To this end, we first conduct a systematic analysis of the RSI power for the PH canceller, which takes into account both the truncation-induced error and the noise-induced error, whereas the former is usually ignored in the existing works. To simplify the performance analysis of RSI power, we employ the generalized Laguerre polynomial (GLP)-based PH canceller instead of the conventional monomial-based one, due to the appealing orthogonality property of the GLP for Gaussian inputs. With the GLP representation of the PH canceller, we further prove that the least-squares channel estimator is asymptotically unbiased, thus demonstrating the asymptotic optimality of Gaussian pilot sequences. Moreover, for the pilot sequence with a finite length, a succinct criterion for minimizing the RSI, namely, the condition-number-to-minimum eigenvalue ratio (CMER) criterion, which essentially balances the truncation-induced and noise-induced error, is presented. By contrast, the existing works normally consider the latter only. Interestingly, it is revealed that an appropriate PAPR of the pilot sequence is of critical importance to achieve the above balance. Simulation results demonstrate that the pilot sequence optimized according to our proposed CMER criterion can achieve an RSI as low as -87.3 dBm, which is over 14 dB lower than that of HE-LTF and over 6 dB lower than that of the state-of-the-art pilot sequence proposed in [1], provided that the order of the PH canceller is no higher than 9 because of the complexity constraint.

eess.SP↗

arXiv2026

Enhanced $s^\pm$-wave superconductivity in electron-doped La$_3$Ni$_2$O$_7$

In cuprates, electron doping yields a much lower superconducting $T_c$ than hole doping. For recently discovered nickelate superconductors, the analogous doping strategies become more challenging. Consequently, while hole-doped Ruddlesden-Popper (RP) nickelates have been extensively studied, electron-doped RP nickelates remain rarely explored both experimentally and theoretically. Here we fill this gap by systematically investigating the two-orbital bilayer model for three representative systems: bulk La$_3$Ni$_2$O$_7$ at ambient pressure and 15\,GPa, and a heterostructure La$_3$Ni$_2$O$_7$:La$_3$Al$_2$O$_7$ that provides a feasible experimental route to electron doping. Using first-principle calculations and large-scale dynamical cluster quantum Monte Carlo simulations, we find that electron doping generically enhances $s^\pm$-wave pairing superconductivity (SC) in all three cases, with the heterostructure showing the highest $T_c$ in the underdoped regime. Furthermore, our results suggest an inter-orbital cooperative mechanism that the pairing on the $d_{x^2-y^2}$ orbital, induced by that on the $d_{z^2}$ orbital, plays a vital role in the SC. This work provides the theoretical prediction of enhanced SC in electron-doped RP nickelates and calls for future experimental verification.

cond-mat.supr-con↗

arXiv2026

EnterpriseRAG: Benchmarking LLM Instruction Adherence and Robustness under Non-Ideal Enterprise Retrieval

Enterprise RAG deployments face a critical reliability gap: while LLMs satisfy 80% of individual constraints, only 26.8% of responses meet all requirements simultaneously, revealing a 57-point orchestration gap. Existing benchmarks assume clean retrieval with simple queries, failing to capture production conditions where noisy documents and multi-dimensional constraints coexist. We introduce EnterpriseRAG, a benchmark of 983 expert-validated samples across six domains that systematically simulates three failure modes absent from prior work: retrieval noise, knowledge gaps, and factual conflicts, coupled with complex instructions. Evaluation of 13 state-of-the-art LLMs reveals a severe instruction adherence collapse, where high per-constraint satisfaction masks low holistic compliance. Critical findings expose deep barriers under knowledge gaps and factual conflicts, even with reasoning-enhanced inference, indicating production RAG requires explicit context-aware protocols and calibrated judgment. EnterpriseRAG provides a reproducible foundation for measuring and closing these gaps, directly informing deployment decisions for enterprise-scale RAG systems. We will release the benchmark and evaluation framework upon publication.

cs.AI↗

arXiv2026

Test-time Adaptation of Pelvic Bone Segmentation Models via Dynamic Reliability-Guided

Reliable pelvic bone segmentation (PBS) from CT is essential for robot-assisted pelvic trauma surgery, yet deploying a source-trained model to a new hospital suffers from severe performance degradation due to cross-center domain shifts. While test-time adaptation (TTA) enables online model adaptation without accessing source data, existing methods show limited effectiveness for PBS, facing challenges including boundary degradation, anatomical inconsistency under domain shifts, and voxel-level class imbalance. To address these challenges, we propose a novel closed-loop dynamic Reliability-Guided TTA framework (ReGA) for PBS. Specifically, we introduce a pseudo-label reliability criterion termed Segmentation Inference Consistency Evaluation (SICE), which jointly measures region overlap and boundary deviation via dropout-based ensemble predictions. Based on SICE, a trust-weighted refinement module adaptively updates features to mitigate boundary errors in pseudo-labels. Furthermore, a confidence-weighted region-level contrastive learning strategy is proposed to enforce anatomical consistency. Finally, ReGA follows the teacher-student (TS) scheme to alleviate voxel-level class imbalance. Experiments on three heterogeneous 3D pelvic CT datasets demonstrate that ReGA consistently outperforms state-of-the-art TTA methods, enabling effective adaptation of the source-trained PBS model to unseen clinical domains. The code is available at https://github.com/Ren-ling/ReGA.

cs.CV↗

arXiv2026

TReB: A Comprehensive Benchmark for Evaluating Table Reasoning Capabilities of Large Language Models

The majority of data in businesses and industries is stored in tables, databases, and data warehouses. Reasoning with table-structured data poses significant challenges for large language models (LLMs) due to its hidden semantics, inherent complexity, and structured nature. One of these challenges is lacking an effective evaluation benchmark fairly reflecting the performances of LLMs on broad table reasoning abilities. In this paper, we fill in this gap by presenting a comprehensive table reasoning benchmark, TReB. Firstly, we propose a taxonomy to systematically measure both shallow table understanding abilities and deep table reasoning abilities, covering a total of 26 sub-tasks. We then construct a high quality dataset through a dedicated data processing and synthesis procedure. Based on these well-constructed samples, we design an evaluation framework to robustly measure table reasoning capabilities with three distinct inference modes. Experimental results with our data and framework reveal that existing LLMs still have significant room for improvement in addressing the complex and real world table related tasks. Both the dataset and evaluation framework are publicly available, with the dataset hosted on https://huggingface.co/datasets/JT-LM/JIUTIAN-TReB, and the framework on https://github.com/JT-LM/jiutian-treb.

cs.CL↗

arXiv2026

Heterostructuring as Gateway to Electron Doping of Nickelate Superconductors

Despite enormous expenditures in the research field, the electron-doped side of nickelate superconductors remains uncharted territory. Substituting the trivalent rare-earth cations by a tetravalent one hitherto failed. Here, we demonstrate by first-principles calculations a disorder-free route to electron dope Ruddlesden-Popper nickelates. When intercalating wide-band-gap insulating layers such as La$X$O$_3$ ($X$=Al, Ga, Sc) into La$_2$NiO$_4$, the extra (LaO)$^+$ layers act as electron donors, releasing carriers into the Ni-3$d$ orbitals. This electron doping puts La$_2$NiO$_4$:La$_2$AlO$_4$ naturally in the optimal region for $d_{x^2-y^2}$-wave superconductivity with T$_c$ exceeding 50 K. The same concept also allows us to electron dope La$_3$Ni$_2$O$_7$, the superconductor in the limelight.

cond-mat.supr-con↗

arXiv2026

DSH-Bench: A Difficulty- and Scenario-Aware Benchmark with Hierarchical Subject Taxonomy for Subject-Driven Text-to-Image Generation

Significant progress has been achieved in subject-driven text-to-image (T2I) generation, which aims to synthesize new images depicting target subjects according to user instructions. However, evaluating these models remains a significant challenge. Existing benchmarks exhibit critical limitations: 1) insufficient diversity and comprehensiveness in subject images, 2) inadequate granularity in assessing model performance across different subject difficulty levels and prompt scenarios, and 3) a profound lack of actionable insights and diagnostic guidance for subsequent model refinement. To address these limitations, we propose DSH-Bench, a comprehensive benchmark that enables systematic multi-perspective analysis of subject-driven T2I models through four principal innovations: 1) a hierarchical taxonomy sampling mechanism ensuring comprehensive subject representation across 58 fine-grained categories, 2) an innovative classification scheme categorizing both subject difficulty level and prompt scenario for granular capability assessment, 3) a novel Subject Identity Consistency Score (SICS) metric demonstrating a 9.4\% higher correlation with human evaluation compared to existing measures in quantifying subject preservation, and 4) a comprehensive set of diagnostic insights derived from the benchmark, offering critical guidance for optimizing future model training paradigms and data construction strategies. Through an extensive empirical evaluation of 19 leading models, DSH-Bench uncovers previously obscured limitations in current approaches, establishing concrete directions for future research and development.

cs.CV↗

arXiv2026

TACO: A Benchmark for Open-Domain Text-to-SQL with Ambiguous and Cross-Database Queries

Text-to-SQL aims to translate natural language questions into executable SQL queries over structured databases. Existing benchmarks mainly focus on closed-domain settings with predefined database schemas and well-specified questions, but they fall short in addressing the challenges of open-domain scenarios, such as ambiguous questions, unspecified databases, and cross-database querying. To bridge this gap, we introduce TACO, a benchmark for open-domain Text-to-SQL with Ambiguous and Cross-database queries. TACO consists of 1,500 real-world Text-to-SQL examples based on a smart city data service and 13,000 high-quality synthetic examples generated based on large-scale open data portals, covering diverse domains such as transportation, healthcare, and finance. To construct the synthetic examples, we develop an effective data synthesis pipeline that preserves the complexity of real-world queries. To demonstrate the utility of TACO, we introduce a baseline TACO-SQL composed of question rewriting, table linking, and query planning, to illustrate the challenges posed by TACO and to better understand the limitations of existing Text-to-SQL approaches. Extensive experiments on TACO using a variety of recent Text-to-SQL approaches show that, while TACO-SQL achieves the best results, a significant gap still remains between the existing approaches and human-written SQL. These findings highlight the difficulty of open-domain Text-to-SQL and position TACO as a valuable benchmark to drive future research.

cs.DB↗

arXiv2026

Intelligent Domain Adaptation for Power System Transient Stability Assessment Under Varying Operating Scenarios

While deep learning-based transient stability assessment (TSA) approaches have exhibited great potential in power system stability monitoring, they are prone to undergo performance degradation in practical contexts with frequent variations of operating conditions. To address this issue, this work develops an adaptive TSA framework via domain adaptation-enabled deep transfer learning. First, for the sake of capturing the primary transient stability characteristics, a robust metric, i.e., heterogeneous hybrid distribution metric (HHDM), is designed through mathematical means to effectively handle multi-scale Gaussian and long-tail distributions of transient responsive data and to precisely quantify the intrinsic distributional discrepancies between the source and target domains corresponding to different operating scenarios. With the help of the HHDM, a Bayesian theory-based dual-distribution domain adaptation method is constructed, aligning not only marginal probability distributions between domains but also the distributions of sub-domain categories. Such alignments enable fine-grained transient stability feature transfer, helping significantly improve the adaptability of a well-trained TSA model to target domains. Furthermore, a multilayer sparse regularization algorithm is introduced to mitigate feature volatility caused by variations in operating scenarios, thereby enhancing the model's generalization in the presence of unforeseen scenarios. Numerical tests on three test systems illustrate that, compared with conventional methods, the proposed framework improves online TSA accuracy by 0.5% to 5% in a cost-effective manner, with the learning cost for TSA model update largely reduced.

eess.SY↗

arXiv2026

Shape Derivatives for Maxwell's Equations with Nonlinear Boundary Conditions

This paper develops a trace-regular variational framework for time-harmonic Maxwell scattering problems involving pointwise nonlinear boundary and interface responses. We investigate three canonical classes of models: nonlinear impedance, nonlinear perfect electric conductor, and nonlinear transmission conditions. Since the standard Maxwell tangential trace belongs to a space of negative order, the nonlinearities are formulated in refined functional spaces where the tangential electric field admits an $L^2(Γ)$-trace. Under the assumption of a sufficiently small Lipschitz constant for the nonlinear response, we establish the well-posedness of the direct problems via fixed-point arguments leveraging the mapping properties of the associated linear Maxwell operators. Within this framework, we perform a rigorous sensitivity analysis of the electromagnetic fields with respect to perturbations of the scattering interface. By employing the covariant Piola transform, we prove the continuity and Fréchet differentiability of the pulled-back solutions with respect to domain variations. The material derivative is characterized as the unique solution to a corresponding $\mathbb{R}$-linearized Maxwell system, and the shape derivative is shown to satisfy explicit boundary or interface conditions for each of the three nonlinear models. We further demonstrate that the resulting sensitivity expressions possess the Hadamard structure, depending exclusively on the normal component of the boundary deformation. The resulting derivative characterizations provide a mathematical basis for subsequent adjoint-based sensitivity analysis, shape optimization, and gradient-driven inverse reconstruction.

math.AP↗

arXiv2026

JT-SAFE-V2: Safety-by-Design Foundation Model with World-Context Data

We introduce JT-Safe-V2, a large language model designed to advance the safety and trustworthiness of foundation models, extending our previous JT-Safe model toward a more comprehensive safety-by-design paradigm. JT-Safe-V2 emphasizes the joint optimization of general intelligence and safety-by-design through several key innovations: enriching pre-training data with contextual world knowledge, high-certainty pre-training procedures, and safety strengthening post-training mechanisms for enterprise-oriented agentic capabilities. Building on these safety-enhanced foundation models, we propose Safe-MoMA (Safe Mixture of Models and Agents), a framework that enables traceable and efficient inference through the orchestrated deployment of multiple models and agents. Extensive evaluations demonstrate that JT-Safe-V2 achieves state-of-the-art performance across both general intelligence and safety benchmarks. Moreover, Safe-MoMA reduces inference costs by more than 30\% compared to using the largest standalone model baseline while maintaining comparable performance. To facilitate future research on safety-by-design foundation models, we publicly release the post-trained JT-Safe-V2-35B model checkpoint.

cs.AI↗

arXiv2026

Adaptive Spatial-Temporal Graph Learning-Enabled Short-Term Voltage Stability Assessment against Time-Varying Topological Conditions

The emerging deep learning (DL) technology has recently exhibited great potential in data-driven short-term voltage stability (SVS) assessment of complex power grids. However, without sufficient attention to the time-varying topological structures of today's power grids, the majority of existing DL-based SVS assessment schemes could experience severe performance degradation in practice. To address this drawback, this paper proposes an adaptive spatial-temporal graph learning-enabled SVS assessment approach that can adapt well to various topological changes. First, considering the time-varying topological conditions of a given power grid, an adaptive graph representation matrix is automatically learned to effectively capture the complicated spatial correlations between individual buses within the grid. Then, to help better capture regional SVS features for subsequent learning processes, the adaptive graph representation matrix is properly adjusted by introducing a spatial attention mechanism. Further, with post-fault system trajectory data linked together via attention-based graph representation, a residual spatiotemporal graph convolutional network is carefully built with Optuna-based optimization to deeply mine system-wide spatiotemporal features and thus achieve structure-adaptive SVS assessment. Numerical test results on two representative sub-systems of a realistic provincial power grid in South China demonstrate the efficacy of the proposed approach under various changing topological conditions.

eess.SY↗