Searcharxiv⌕ Search

arXiv subjects

Wang Yang

Publications and source records attributed to Wang Yang.

At least 19 recordsLinked to original sources
arXiv2026

Towards Trustworthy Physical AI: From Theory to Practice Across Life Cycle

Physical AI refers to AI systems that understand, reason about, and act in accordance with the physical world and its underlying laws, dynamics, and constraints. Unlike conventional AI systems, physical AI interacts continuously with uncertain physical environments, and its actions produce consequences that are physically irreversible. As existing trustworthy AI frameworks have been developed primarily for digital AI systems, they do not fully capture the distinctive challenges of physical AI, such as physical safety, cyber-physical security, and physical manufacturing process. To address this gap, we present a survey of trustworthy physical AI principles. First, we characterize the core capabilities and challenges of physical AI. Second, we examine the role of physics in AI. Third, we trace the end-to-end physical AI life cycle across five core stages and introduce Trustworthy Physical AI Operationalization (T-PAIO). Fourth, we develop the Trustworthy Physical AI (T-PAI) framework, a theoretical framework that organizes key trustworthiness principles and provides a foundation for governing trustworthy physical AI systems.

cs.AI↗
arXiv2026

Crossover and Changeover in Spin-1 Kitaev-$\Gamma$ Chain with Uniaxial Single-ion Anisotropy

Recent advances in bond-directional spin chains have revealed extensive emergent phenomena and unconventional criticality. Here we investigate the spin-1 Kitaev-$\Gamma$ chain with uniaxial single-ion anisotropy (SIA) using large-scale density-matrix renormalization group calculations and bosonization analysis. Tuning the SIA strength reveals a crossover from the Kitaev phase to the large-$D$ phase, evidenced by the excitation gap changing from quadratic to linear, the coexistence and smooth evolution of spin-nematic and string order parameters, and the suppression of the double-peak specific heat. For negative SIA, we uncover a changeover from a first-order transition to a continuous one between the dimerized and Haldane phases. The continuous transition belongs to the \textrm{SU(2)$_2$} Wess-Zumino-Witten universality class with central charge $c=3/2$, a rare instance in a system without continuous symmetry. Our results establish the Kitaev-$\Gamma$ chain as a minimal platform for controlling crossover and changeover phenomena.

cond-mat.str-el↗
arXiv2026

Mechanistic Diagnostics of Spatial Lexical Bias in Multimodal Large Language Model Spatial Reasoning

Multimodal large language models (MLLMs) remain unreliable on spatial multiple-choice questions, and their failures are often attributed to poorly attended visual information. We identify a complementary failure mode, spatial lexical bias: a spatial relation word added to the answer options can act as a lexical-semantic distractor that draws the model's decision toward that option. Using nine open-weight MLLMs, we show that this phenomenon is widespread. We then isolate diagnostic cases in which a model answers a binary spatial question correctly yet consistently chooses a newly added third spatial option, which we call binary-stable but ternary-fragile cases. Leveraging mechanistic interpretability tools on these cases, we find that the failure arises on the language side rather than the visual side: visual attention analyses and residual-stream probes show the correct spatial relation remains internally available, while irrelevant-option controls, activation patching, and sparse component interventions trace the bias to specific LLM-side channels and neurons. Accordingly, we show that a lightweight LLM-only DPO update on tiny single-object-pair synthetic data mitigates the bias, lifting four-way robust accuracy by up to 100 points on synthetic data, and by 68.0, 32.6, and 20.1 points on broader evaluation datasets WhatsUp, SpatialMQA-Direct, and VSR.

cs.CL↗
arXiv2026

Renormalization group analysis for bosonization coefficients in half-odd-integer Kitaev spin chains

Based on a renormalization group (RG) analysis, we study the bosonization formulas in spin-$S$ Kitaev-Gamma and Kitaev-Heisenberg-Gamma chains in the $(K<0,\Gamma>0,J>0)$ parameter region, where $S$ is a half-odd integer. We find that the effects associated with the breaking of emergent continuous symmetries in bosonization formulas scale as $1/S$ in the large-$S$ limit, which is in qualitative agreement with DMRG numerical results for Kitaev-Gamma chains. In Kitaev-Heisenberg-Gamma chains, symmetry analysis reveals ten independent bosonization coefficients, five of which are predicted by the RG analysis to have no dependence on the Heisenberg coupling up to linear order. Our work may offer valuable input for determining magnetic ordering tendencies in two-dimensional Kitaev spin models within a quasi-one-dimensional approach.

cond-mat.str-el↗
arXiv2026

Path-Lock Expert: Separating Reasoning Mode in Hybrid Thinking via Architecture-Level Separation

Hybrid-thinking language models expose explicit /think and /no_think modes, but current designs do not separate them cleanly. Even in /no_think mode, models often emit long and self-reflective responses, causing reasoning leakage. Existing work reduces this issue through better data curation and multi-stage training, yet leakage remains because both modes are still encoded in the same feed-forward parameters. We propose Path-Lock Expert (PLE), an architecture-level solution that replaces the single MLP in each decoder layer with two semantically locked experts, one for /think and one for /no_think, while keeping attention, embeddings, normalization, and the language-model head shared. A deterministic control-token router selects exactly one expert path for the entire sequence, so inference preserves the dense model's per-token computation pattern and each expert receives mode-pure updates during supervised fine-tuning. Across math and science reasoning benchmarks, PLE maintains strong /think performance while producing a substantially stronger mode separation, a /no_think mode with higher accuracy and far less reasoning leakage. On Qwen3-4B, for example, compared to the SFT-only baseline on AIME24, PLE generates 17x fewer reflective tokens (6.01 vs. 0.35 per answer) and 2x shorter outputs (8665 vs. 4101 tokens), and improves /no_think accuracy from 35.33% to 44.67%, while maintaining /think-mode performance (61.33% vs. 60.00%). These results suggest that controllable hybrid thinking is fundamentally an architectural problem, and separating mode-specific feed-forward pathways is a simple and effective solution.

cs.CL↗
arXiv2026

Topological multicomponent superconductivity with sizable $s$-wave admixture in twisted bilayer cuprates

We investigate multicomponent superconductivity in twisted bilayer cuprates with order parameter $s+d_1 e^{i\phi_1}+d_2 e^{i\phi_2}$, where $s=s_1+s_2$ is the symmetric layer-resolved $s$-wave component and $d_i$ denotes the $d$-wave pairing in layer $i$. When $\phi_1-\phi_2\neq 0,\pi$, this three-component state breaks time-reversal and $C_4$ rotational symmetries and is topologically nontrivial. Combining Ginzburg--Landau analysis with self-consistent microscopic mean-field calculations, we show that this topological state is stabilized over a broad parameter regime. We further identify nematic Kerr anisotropy as a smoking-gun signature distinguishing it from $s+id$ and $d_1+e^{i\phi}d_2$ states. Our results show that a sizable $s$-wave component does not preclude chiral topological superconductivity, pointing to twisted cuprates as a more robust platform than previously appreciated.

cond-mat.supr-con↗
arXiv2026

AgentCE-Bench: Agent Configurable Evaluation with Scalable Horizons and Controllable Difficulty under Lightweight Environments

Existing Agent benchmarks suffer from two critical limitations: high environment interaction overhead (up to 41\% of total evaluation time) and imbalanced task horizon and difficulty distributions that make aggregate scores unreliable. To address these issues, we propose AgentCE-Bench built around a unified grid-based planning task, where agents must fill hidden slots in a partially completed schedule subject to both local slot constraints and global constraints. Our benchmark offers fine-grained control through two orthogonal axes: \textbf{Scalable Horizons}, controlled by the number of hidden slots $H$, and \textbf{Controllable Difficulty}, governed by a decoy budget $B$ that determines the number of globally misleading decoy candidates. Crucially, all tool calls are resolved via static JSON files under a \textbf{Lightweight Environment} design, eliminating setup overhead and enabling fast, reproducible evaluation suitable for training-time validation. We first validate that $H$ and $B$ provide reliable control over task horizon and difficulty, and that AgentCE-Bench exhibits strong domain consistency and model discriminability. We then conduct comprehensive experiments across 13 models of diverse sizes and families over 6 domains, revealing significant cross-model performance variation and confirming that AgentCE-Bench provides interpretable and controllable evaluation of agent reasoning.

cs.AI↗
arXiv2026

Theoretical study of spin-dependent transport in WSe$_2$-based vertical spin valves

We theoretically investigate spin-dependent transport in a TMD-based vertical spin valve, taking WSe$_2$ as a representative example. Using effective Hamiltonians for the heterostructure and the Landauer formula, we derive the transmission and reflection coefficients within a transfer-matrix approach. The calculated magnetoresistance shows an oscillatory dependence on the WSe$_2$ thickness when the Fermi level is tuned near the valence-band maximum. The effects of gate voltage and exchange fields on the magnetoresistance are further analyzed. We also identify a Fabry-P\'erot-like interference contribution to the magnetoresistance, which can enhance or even induce negative magnetoresistance in certain thickness regimes. Our results provide a qualitative understanding of the negative magnetoresistance observed in WSe$_2$-based spin valves and may offer useful insights for the design of tunable spintronic devices.

cond-mat.mes-hall↗
arXiv2026

MOSAIC: Multi-Objective Slice-Aware Iterative Curation for Alignment

We study how to allocate a fixed supervised fine-tuning budget when three objectives must be balanced at once: multi-turn safety alignment, low over-refusal on benign boundary queries, and instruction following under verifiable constraints. We propose MOSAIC (Multi-Objective Slice-Aware Iterative Curation for Alignment), a multi-objective framework for closed-loop data mixture search built on a unified L1-L3 evaluation interface. MOSAIC turns slice-level failure profiles into executable data actions, including dataset-level mixture ratios, bucket-level weights, and focus criteria. Under a fixed 1M-token budget and five rounds of independent fine-tuning from the same base model, MOSAIC improves internal XGuard from 2.76 to 4.67 while keeping OrBench at 4.41 and IFEval at 3.65. The final Pareto solution also generalizes better than a random static LoRA baseline on independent attack, over-refusal, and capability tests, suggesting that structured failure diagnosis can serve as a practical control signal for budgeted data construction. Code is available at https://github.com/douyipu/mosaic.

cs.CR↗
arXiv2026

Toward Trustworthy Evaluation of Sustainability Rating Methodologies: A Human-AI Collaborative Framework for Benchmark Dataset Construction

Sustainability or ESG rating agencies use company disclosures and external data to produce scores or ratings that assess the environmental, social, and governance performance of a company. However, sustainability ratings across agencies for a single company vary widely, limiting their comparability, credibility, and relevance to decision-making. To harmonize the rating results, we propose adopting a universal human-AI collaboration framework to generate trustworthy benchmark datasets for evaluating sustainability rating methodologies. The framework comprises two complementary parts: STRIDE (Sustainability Trust Rating & Integrity Data Equation) provides principled criteria and a scoring system that guide the construction of firm-level benchmark datasets using large language models (LLMs), and SR-Delta, a discrepancy-analysis procedural framework that surfaces insights for potential adjustments. The framework enables scalable and comparable assessment of sustainability rating methodologies. We call on the broader AI community to adopt AI-powered approaches to strengthen and advance sustainability rating methodologies that support and enforce urgent sustainability agendas.

cs.AI↗
arXiv2026

When Domains Interact: Asymmetric and Order-Sensitive Cross-Domain Effects in Reinforcement Learning for Reasoning

Group Relative Policy Optimization (GRPO) has become a key technique for improving reasoning abilities in large language models, yet its behavior under different domain sequencing strategies is poorly understood. In particular, the impact of sequential (one domain at a time) versus mixed-domain (multiple domain at a time) training in GRPO has not been systematically studied. We provide the first systematic analysis of training-order effects across math, science, logic, and puzzle reasoning tasks. We found (1) single-domain generalization is highly asymmetric: training on other domains improves math reasoning by approximately 25\% accuracy, while yielding negligible transfer to logic and puzzle; (2) cross-domain interactions are highly order-dependent: training in the order math$\rightarrow$science achieves 83\% / 41\% accuracy on math / science, while reversing the order to science$\rightarrow$math degrades performance to 77\% / 25\%; (3) no single strategy is universally optimal in multi-domain training: sequential training favors math (up to 84\%), mixed training favors science and logic, and poor ordering can incur large performance gaps (from 70\% to 56\%). Overall, our findings demonstrate that GRPO under multi-domain settings exhibits pronounced asymmetry, order sensitivity, and strategy dependence, highlighting the necessity of domain-aware and order-aware training design.

cs.LG↗
arXiv2026

AJAR: Adaptive Jailbreak Architecture for Red-teaming

Large language model (LLM) safety evaluation is moving from content moderation to action security as modern systems gain persistent state, tool access, and autonomous control loops. Existing jailbreak frameworks still leave a gap between adaptive multi-turn attacks and agentic runtimes: attack algorithms are usually packaged as monolithic scripts, while agent harnesses rarely expose explicit abstractions for rollback, tool simulation, or strategy switching. We present AJAR, a red-teaming framework that exposes multi-turn jailbreak algorithms as callable MCP services and lets an Auditor Agent orchestrate them inside a tool-aware runtime built on Petri. AJAR integrates three representative attacks, namely Crescendo, ActorAttack, and X-Teaming, under a shared service interface for planning, prompt generation, optimization, evaluation, and context control. On 200 HarmBench validation behaviors, AJAR improves X-Teaming from 65.0% to 76.0% attack success rate (ASR), reaches 80% cumulative success one turn earlier than the native implementation, and reproduces Crescendo more effectively than PyRIT (91.0% vs. 87.5% ASR). Behavior-level analysis shows that these gains are concentrated in hard categories and frequently depend on rollback-enabled transcript repair. We further show that tool access reshapes rather than uniformly enlarges the attack surface: ActorAttack rises from 51.0% to 56.0% ASR with tools, whereas Crescendo drops from 91.0% to 78.0% and X-Teaming from 76.0% to 55.5%, with the sharpest declines appearing in categories that rely on long semantic buildup. These results position AJAR as a practical foundation for evaluating multi-turn jailbreaks under realistic agent constraints. Code and data are available at https://github.com/douyipu/ajar.

cs.CR↗
arXiv2026

Electric field effects in one-dimensional spin-1/2 $K_1J_1\Gamma_1\Gamma_1^\prime K_2J_2$ model with ferromagnetic Kitaev coupling

We perform a systematic study on the effects of electric fields in the Luttinger liquid phase of the one-dimensional spin-$1/2$ $K_1J_1\Gamma_1\Gamma_1^\prime K_2J_2$ model in the region of ferromagnetic nearest-neighboring Kitaev coupling. We find that while electric fields along $(1,1,1)$-direction maintain the Luttinger liquid behavior, fields along other directions drive the system to a dimerized state. An estimation is made on how effective a $(1,1,1)$-field is for tuning the Luttinger parameter in real materials. Our work is useful for understanding the effects of electric fields in one-dimensional generalized Kitaev spin models, and provides a starting point for exploring the electric-field-related physics in two dimensions based on a quasi-one-dimensional approach.

cond-mat.str-el↗
arXiv2026

Mid-Think: Training-Free Intermediate-Budget Reasoning via Token-Level Triggers

Hybrid reasoning language models are commonly controlled through high-level Think/No-think instructions to regulate reasoning behavior, yet we found that such mode switching is largely driven by a small set of trigger tokens rather than the instructions themselves. Through attention analysis and controlled prompting experiments, we show that a leading ``Okay'' token induces reasoning behavior, while the newline pattern following `` '' suppresses it. Based on this observation, we propose Mid-Think, a simple training-free prompting format that combines these triggers to achieve intermediate-budget reasoning, consistently outperforming fixed-token and prompt-based baselines in terms of the accuracy-length trade-off. Furthermore, applying Mid-Think to RL training after SFT reduces training time by approximately 15% while improving final performance of Qwen3-8B on AIME from 69.8% to 72.4% and on GPQA from 58.5% to 61.1%, demonstrating its effectiveness for both inference-time control and RL-based reasoning training.

cs.CL↗
arXiv2025

Many-body wave function and edge magnetization of an open $p+is$ superconducting chain

Although BCS wave functions for superconductors under periodic boundary conditions are well established, obtaining an explicit form of the many-body BCS wave function under open boundary condition is usually a nontrivial problem. In this work, we construct the exact BCS ground-state wave function of a one-dimensional spin-$\frac12$ superconductor with $p+is$ pairing symmetry under open boundary conditions for special sets of parameters. The spin magnetization on the edges is calculated explicitly using the obtained wave function. Our work is useful for obtaining deeper understandings of open $p+ is$ superconducting chains on a wave-function level.

cond-mat.supr-con↗
arXiv2025

Demystifying Hybrid Thinking: Can LLMs Truly Switch Between Think and No-Think?

Hybrid thinking enables LLMs to switch between reasoning and direct answering, offering a balance between efficiency and reasoning capability. Yet our experiments reveal that current hybrid thinking LLMs only achieve partial mode separation: reasoning behaviors often leak into the no-think mode. To understand and mitigate this, we analyze the factors influencing controllability and identify four that matter most: (1) larger data scale, (2) using think and no-think answers from different questions rather than the same question, (3) a moderate increase in no-think data number, and (4) a two-phase strategy that first trains reasoning ability and then applies hybrid think training. Building on these findings, we propose a practical recipe that, compared to standard training, can maintain accuracy in both modes while significantly reducing no-think output length (from $1085$ to $585$ on MATH500) and occurrences of reasoning-supportive tokens such as ``\texttt{wait}'' (from $5917$ to $522$ on MATH500). Our findings highlight the limitations of current hybrid thinking and offer directions for strengthening its controllability.

cs.LG↗
arXiv2025

Gapless and ordered phases in spin-1/2 Kitaev-XX-Gamma chain

In this work, we study the spin-1/2 Kitaev chain with additional XX and symmetric off-diagonal Gamma interactions. By a combination of Jordan-Wigner transformation and density matrix renormalization group (DMRG) numerical simulations, we obtain the exact solution of the model and map out the phase diagram containing six distinct phases. The four gapped phases display ferromagnetic and antiferromagnetic magnetic orders along the (1, 1, 0)- and (1, -1, 0)-spin directions, whereas in the gapless phases, the low energy spectrum consists of two branches of helical Majorana fermions with unequal velocities. Transition lines separating different phases include deconfined quantum critical lines with dynamical critical exponent z = 1 and quadratic critical lines with z = 2. Our work reveals the rich interplay among symmetry, magnetic order, and quantum criticality in the Kitaev-XX-Gamma chain

cond-mat.str-el↗
arXiv2025

100-LongBench: Are de facto Long-Context Benchmarks Literally Evaluating Long-Context Ability?

Long-context capability is considered one of the most important abilities of LLMs, as a truly long context-capable LLM enables users to effortlessly process many originally exhausting tasks -- e.g., digesting a long-form document to find answers vs. directly asking an LLM about it. However, existing real-task-based long-context evaluation benchmarks have two major shortcomings. First, benchmarks like LongBench often do not provide proper metrics to separate long-context performance from the model's baseline ability, making cross-model comparison unclear. Second, such benchmarks are usually constructed with fixed input lengths, which limits their applicability across different models and fails to reveal when a model begins to break down. To address these issues, we introduce a length-controllable long-context benchmark and a novel metric that disentangles baseline knowledge from true long-context capabilities. Experiments demonstrate the superiority of our approach in effectively evaluating LLMs.

cs.CL↗