Searcharxiv⌕ Search

arXiv subjects

Zhiyuan Sun

Publications and source records attributed to Zhiyuan Sun.

At least 19 recordsLinked to original sources
arXiv2026

F$^{2}$DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch Workflows

With the widespread industrial deployment of Large Language Models (LLMs), DeepSearch has emerged as the dominant paradigm for resolving complex user queries. It typically operates through an iterative closed-loop workflow consisting of planning and reflection, information retrieval, and answer generation. However, existing reward models (RMs) and evaluation benchmarks are primarily designed for static single-turn tasks, failing to capture the full-pipeline complexity of DeepSearch workflows. To address this limitation, we propose F2DR, a fine-grained full-pipeline DeepSearch reward framework. F2DR evaluates DeepSearch workflows across three dimensions: Content, Trajectory, and Answer, enabling comprehensive process-level assessment. We further construct DeepSearch RM-Bench, a dedicated benchmark for evaluating RMs in DeepSearch scenarios. Extensive experiments demonstrate that F2DR achieves significantly higher evaluation consistency than self-evaluation-based baselines, while DeepSearch RM-Bench exhibits strong discriminative capability across existing open-source RMs. We will publicly release the complete DeepSearch RM-Bench dataset soon.

cs.CL↗
arXiv2026

A Thermodynamically Consistent High-Order Framework for Staggered Lagrangian Hydrodynamics

We present a consistent high-order staggered Lagrangian hydrodynamics framework designed to reconcile an underlying disparity in existing curvilinear formulations: the mismatch between quadrature-based "strong" mass conservation and the discrete degrees of freedom (DOFs) of thermodynamic variables. By mathematically coupling the numerical quadrature rule with the density representation, our approach ensures rigorous point-wise consistency between density, internal energy, and pressure. This synchronization eliminates the ambiguity of equation-of-state (EOS) updates inherent in previous high-order staggered methods. To stabilize the discretization, we develop a high-order generalization of the subzonal pressure method by conceptually enriching the pressure field from the $Q^{m-1}$ to the $Q^m$ finite element space. We prove that evaluating this enriched field using a high-order quadrature rule naturally generates a restorative anti-hourglass force, which exactly recovers the classical $Q^1-P^0$ compatible hydrodynamics algorithm as a limiting case for $m=1$. Furthermore, we introduce a concise, algorithmic formulation of tensor artificial viscosity that streamlines implementation and significantly reduces computational overhead in high-order settings. The resulting framework yields strictly diagonal mass matrices for both momentum and energy equations, enabling highly efficient, fully explicit time integration without global linear solves. Extensive numerical benchmarks, including smooth convergence tests and complex shock-dominated flows, demonstrate that the proposed method achieves optimal high-order accuracy while maintaining superior geometric robustness.

math.NA↗
arXiv2026

Light-induced nonconservative static forces in many-body systems

In a quantum many-body system, a periodic drive can often generate effective static forces on the slow collective degrees of freedom. We study the static forces generated by light on electronic order parameters in solid-state systems. We show that the forces can have nonconservative components that originate from dissipation, i.e., optical absorption. This effect is demonstrated in two nontrivial examples. In excitonic insulators, via interband excitations, the light field generates a nonconservative force on the phase of the excitonic order parameter. This force leads to an acceleration of the phase, which manifests as a shift of the photon emission peak from an exciton condensate. In materials with an incommensurate charge density wave, a propagating light field generates a nonconservative force on the phase of its order parameter. It drives the charge density wave into sliding motion, leading to a DC electric current via topological Thouless pumping.

cond-mat.stat-mech↗
arXiv2026

Floquet polaritons in optically driven materials

Polaritons are coupled collective modes of light and matter in quantum materials. In modern pump-probe experiments, a pump light pulse may dramatically alter the properties of the polaritons, rendering them Floquet polaritons that can be detected by a probe pulse. We present a practical framework to describe Floquet polaritons in terms of the linear and nonlinear optical properties of the material. The central quantity that yields the spectra of Floquet polaritons is an effective linear optical susceptibility contributed by the pump through nonlinear optical susceptibilities. We apply this method to graphene and show that via its third-order optical nonlinearity, infrared pump leads to Floquet plasmon bands. Notably, near plasmonic band crossings, parametric instability leads to flat bands with unstable modes and exceptional points that closely resemble those of non-Hermitian systems. As a second example, we show that in hexagonal boron nitride pumped by mid-infrared laser, the pump induces Floquet phonon polariton bands via phononic nonlinearity, which can be detected with either far-field or near-field optical technique. Finally, in layered superconductors pumped by THz light polarized along the out-of-plane direction, the Josephson-type optical nonlinearity leads to Floquet Josephson plasmons, which manifest as new peaks in the THz reflectivity of a probe pulse.

cond-mat.mes-hall↗
arXiv2026

Thinking as Compression: Your Reasoning Model is Secretly a Context Compressor

Context compression aims to shorten long context inputs with minimal information loss for LLM inference acceleration. While existing methods have shown promise, they typically rely on complex compression modules or compression-specific training, leaving the intrinsic capabilities of LLMs underexplored. In contrast, this work reveals that a thinking model itself can naturally compress long contexts by organizing task-relevant information. We thus derive Thinking as Compression (TaC), a new compression paradigm that treats thinking itself as compressed context. Without relying on specific dedicated compressor, TaC directly prompts the thinking model to generate thinking traces as the shortened context, already outperforming most representative compression methods. Further, given that raw thinking output may struggle with budget control and shortcut behaviors, we introduce Thinking as Compression Constrained (TaC-C), leveraging a simple reward-driven optimization framework to elicit intrinsic thinking as compact and controllable compressed context. Experiments across four long-context QA benchmarks demonstrate that TaC-C consistently outperforms existing baselines. At 4x and 8x compression ratios, it surpasses the strongest competitor by 17.4% and 23.4% in average F1, and by 15.7% and 21.7% in average Exact Match Score (EM), respectively.

cs.AI↗
arXiv2026

When Less is More: The LLM Scaling Paradox in Context Compression

Scaling up model parameters has long been a prevalent training paradigm driven by the assumption that larger models yield superior generation capabilities. However, under lossy context compression in a compressor--decoder setup, we find a \textbf{\textit{Size-Fidelity Paradox}}: increasing compressor size can lessen the faithfulness of reconstructed contexts though reconstruction error decreases. Across 27 compressor setups spanning model families, scales, and compression rates, we coin this paradox arising from two dominant factors: 1) \textit{knowledge overwriting}: larger models increasingly replace source facts with their own prior beliefs, \textit{e.g.}, ``the white strawberry`` $\to$ ``the red strawberry``; and 2) \textit{semantic drift}: larger models tend to paraphrase or restructure content instead of reproducing it verbatim, \textit{e.g.}, ``Alice hit Bob`` $\to$ ``Bob hit Alice``. Interestingly, this paradox persists across varied settings, with mid-sized compressors often outperforming larger ones in faithful recovery. By analyzing the compressed memory via embedding geometry and reconstruction determinacy, we further reveal that compressors tend to organize memory across broader semantic subspaces, yielding more ambiguous representations prone to overwriting, drift, and weakened recovery. These findings complement existing evaluations of context compression and expose a breakdown of scaling laws when the objective shifts from plausible generation to faithful preservation.

cs.LG↗
arXiv2026

ConsistRM: Improving Generative Reward Models via Consistency-Aware Self-Training

Generative reward models (GRMs) have emerged as a promising approach for aligning Large Language Models (LLMs) with human preferences by offering greater representational capacity and flexibility than traditional scalar reward models. However, GRMs face two major challenges: reliance on costly human-annotated data restricts scalability, and self-training approaches often suffer from instability and vulnerability to reward hacking. To address these issues, we propose ConsistRM, a self-training framework that enables effective and stable GRM training without human annotations. ConsistRM incorporates the Consistency-Aware Answer Reward, which produces reliable pseudo-labels with temporal consistency, thereby providing more stable model optimization. Moreover, the Consistency-Aware Critique Reward is introduced to assess semantic consistency across multiple critiques and allocates fine-grained and differentiated rewards. Experiments on five benchmark datasets across four base models demonstrate that ConsistRM outperforms vanilla Reinforcement Fine-Tuning (RFT) by an average of 1.5%. Further analysis shows that ConsistRM enhances output consistency and mitigates position bias caused by input order, highlighting the effectiveness of consistency-aware rewards in improving GRMs. Our implementation is available at https://github.com/yuliangCarmelo/ConsistRM.

cs.AI↗
arXiv2026

ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework

Reward Models (RMs) are critical components in the Reinforcement Learning from Human Feedback (RLHF) pipeline, directly determining the alignment quality of Large Language Models (LLMs). Recently, Generative Reward Models (GRMs) have emerged as a superior paradigm, offering higher interpretability and stronger generalization than traditional scalar RMs. However, existing methods for GRMs focus primarily on outcome-level supervision, neglecting analytical process quality, which constrains their potential. To address this, we propose ReflectRM, a novel GRM that leverages self-reflection to assess analytical quality and enhance preference modeling. ReflectRM is trained under a unified generative framework for joint modeling of response preference and analysis preference. During inference, we use its self-reflection capability to identify the most reliable analysis, from which the final preference prediction is derived. Experiments across four benchmarks show that ReflectRM consistently improves performance, achieving an average accuracy gain of +3.7 on Qwen3-4B. Further experiments confirm that response preference and analysis preference are mutually reinforcing. Notably, ReflectRM substantially mitigates positional bias, yielding +10.2 improvement compared with leading GRMs and establishing itself as a more stable evaluator. Our code is available at https://github.com/yuliangCarmelo/ReflectRM.

cs.AI↗
arXiv2026

Automated Functional Testing for Malleable Mobile Application Driven from User Intent

Software malleability allows applications to be easily changed, configured, and adapted even after deployment. While prior work has explored configurable systems, adaptive recommender systems, and malleable GUIs, these approaches are often tailored to specific software and lack generalizability. In this work, we envision per-user malleable mobile applications, where end-users can specify requirements that are automatically implemented via LLM-based code generation. However, realizing this vision requires overcoming the key challenge of designing automated test generation that can reliably verify both the presence and correctness of user-specified functionalities. We propose ALADDIN, a user-requirement-driven GUI test generation framework that incrementally navigates the UI, triggers desired functionalities, and constructs LLM-guided oracles to validate correctness. We build a benchmark spanning six popular mobile applications with both correct and faulty user-requested functionalities, demonstrating that ALADDIN effectively validates per-user features and is practical for real-world deployment. Our work highlights the feasibility of shifting mobile app development from a product-manager-driven to an end-user-driven paradigm.

cs.SE↗
arXiv2026

Supercurrent Growth in Nonequilibrium Superconductors

In ultrafast experiments on superconductors, a pump laser pulse often heats up the electronic system and suppresses the density of superfluid electrons. Subsequently, the electrons undergo a cooling process because of electron-phonon thermalization so that the superfluid density recovers in time. We study the nonequilibrium electromagnetic response of the system in this cooling process. We show that if a supercurrent is initiated by a probe electric field pulse, an intriguing phenomenon of `supercurrent growth' occurs, meaning that the net current grows in time with the increasing superfluid density. Using the Boltzmann kinetic equation, we uncover its microscopic origin as the momentum-relaxing scattering of Bogoliubov quasiparticles by impurities and phonons, in stark contrast to the widely accepted intuition that impurities always attenuate currents. We further show that supercurrent growth has important experimental manifestations, including the ultrafast Meissner effect and an optical reflectivity exceeding unity.

cond-mat.supr-con↗
arXiv2026

Advancing General-Purpose Reasoning Models with Modular Gradient Surgery

Reinforcement learning (RL) has played a central role in recent advances in large reasoning models (LRMs), yielding strong gains in verifiable and open-ended reasoning. However, training a single general-purpose LRM across diverse domains remains challenging due to pronounced domain heterogeneity. Through a systematic study of two widely used strategies, Sequential RL and Mixed RL, we find that both incur substantial cross-domain interference at the behavioral and gradient levels, resulting in limited overall gains. To address these challenges, we introduce **M**odular **G**radient **S**urgery (**MGS**), which resolves gradient conflicts at the module level within the transformer. When applied to Llama and Qwen models, MGS achieves average improvements of 4.3 (16.6\%) and 4.5 (11.1\%) points, respectively, over standard multi-task RL across three representative domains (math, general chat, and instruction following). Further analysis demonstrates that MGS remains effective under prolonged training. Overall, our study clarifies the sources of interference in multi-domain RL and presents an effective solution for training general-purpose LRMs.

cs.CL↗
arXiv2026

TRE: Encouraging Exploration in the Trust Region

Entropy regularization is a standard technique in reinforcement learning (RL) to enhance exploration, yet it yields negligible effects or even degrades performance in Large Language Models (LLMs). We attribute this failure to the cumulative tail risk inherent to LLMs with massive vocabularies and long generation horizons. In such environments, standard global entropy maximization indiscriminately dilutes probability mass into the vast tail of invalid tokens rather than focusing on plausible candidates, thereby disrupting coherent reasoning. To address this, we propose Trust Region Entropy (TRE), a method that encourages exploration strictly within the model's trust region. Extensive experiments across mathematical reasoning (MATH), combinatorial search (Countdown), and preference alignment (HH) tasks demonstrate that TRE consistently outperforms vanilla PPO, standard entropy regularization, and other exploration baselines. Our code is available at https://github.com/WhyChaos/TRE-Encouraging-Exploration-in-the-Trust-Region.

cs.CL↗
arXiv2026

Electrically tunable dipolar polaritons with giant nonlinearity in a homobilayer microcavity

Active control over strong optical nonlinearity in solid-state systems is central to unlocking exotic many-body phenomena and scalable photonic devices. While exciton-polaritons in transition metal dichalcogenides (TMDs) offer a promising platform, their practical utility is often impeded by fixed interaction parameters and an intrinsic trade-off between nonlinearity and oscillator strength. Here, we report electrically tunable dipolar polaritons in a dual-gated bilayer MoS2 microcavity, demonstrating in situ reshaping of the dispersion and modulation of the light-matter coupling strength via the quantum-confined Stark effect. Crucially, this architecture enables a giant polariton-polariton interaction strength tunable by a factor of seven. This nonlinearity enhancement arises from a synergistic interplay, in which the electric field amplifies the microscopic dipolar repulsion while simultaneously optimizing the macroscopic excitonic Hopfield coefficient. Furthermore, electrostatic doping serves as an independent control knob to switch the system between strong and weak coupling regimes. Our findings bridge the gap between strong optical coupling and giant dipolar nonlinearities, establishing the TMD homobilayer as a versatile platform for engineering programmable correlated many-body states on a chip.

physics.optics↗
arXiv2026

Universal Phase Transitions of Matter in Optically Driven Cavities

Optical cavities have been widely applied to manipulate the properties of solid state materials inside them. We propose that in systems embedded within optical cavities driven by incident pump light, the pump induces generic phase transitions into new nonequilibrium steady states. This effect arises from the ponderomotive potential, the effective static potential exerted by the pump on the low energy degrees of freedom, which exhibits a universal steplike structure that pushes the matter degrees of freedom in the direction that redshifts the cavity photon modes. For a two dimensional electron liquid in a driven cavity, this steplike potential pushes the electron density to jump to a smaller value so that a hybrid cavity photon mode is redshifted to slightly below the pump frequency. Similarly, for a dirty superconductor in such a driven cavity, this potential acts on the superconducting order parameter and causes a first order phase transition to a new steady state with a smaller gap. By realistic electromagnetic modeling of the cavity that includes all cavity modes, we construct the nonequilibrium phase diagrams for experimentally relevant devices.

cond-mat.mes-hall↗
arXiv2026

Optical Detection and Manipulation of Pseudospin Orders in Wigner Crystals

In Wigner-crystal states of two-dimensional electrons, the spin ordering remains poorly understood. The small energy differences between candidate spin orders make theoretical studies less reliable, and probing magnetic order at a nonzero wave vector is experimentally challenging. In modern realizations of Wigner crystals, the electronic spin degree of freedom is often replaced by a valley pseudospin associated with nonzero Berry curvature. The resulting anomalous velocity couples the electrons' pseudospin texture to their orbital vibration. We show that this mechanism enables optical detection of pseudospin orders in Wigner crystals by producing sharp signatures in the terahertz optical conductivity. For example, antiferromagnetic pseudospin order enables light to excite collective electronic vibrations at the ordering wave vector, generating a characteristic absorption peak. Based on the same principle, we further show that a strong optical drive generates an effective potential that reshapes the pseudospin energy landscape, inducing phase transitions to stripe antiferromagnetic states. These results point to a route for optical detection and control of spin order via its coupling to orbital motion.

cond-mat.str-el↗
arXiv2025

Ab initio Approach to Collective Excitations in Excitonic Insulators

An ab initio approach is presented for studying the collective excitations in excitonic insulators, charge/spin density waves and superconductors. We derive the Bethe-Salpeter-Equation for the particle-hole excitations in the quasiparticle representation, from which the collective excited states are solved and the corresponding order parameter fluctuations are computed. This method is demonstrated numerically for the excitonic insulating phases of the biased WSe2-MoSe2 bilayer. It reveals the gapless phase-mode, the subgap Bardasis-Schrieffer modes and the above-gap scattering states. Our work paves the way for quantitative predictions of excited state phenomena from first-principles calculations in electronic systems with spontaneous symmetry breaking.

cond-mat.mtrl-sci↗
arXiv2025

Floquet engineering enabled by charge density wave transition

Floquet engineering has emerged as a powerful approach for dynamically tailoring the electronic structures of quantum materials through time-periodic light fields generated by ultrafast laser pulses. The light fields can transiently dress Bloch electrons, creating novel electronic states inaccessible in equilibrium. While such temporal modulation provides dynamic control, spatially periodic modulations, such as those arising from charge density wave (CDW) order, can also dramatically reconstruct the band structure through real-space symmetry breaking. The interplay between these two distinct forms of modulation-temporal and spatial-opens a new frontier in electronic-phase-dependent Floquet engineering. Here we demonstrate this concept experimentally in the prototypical CDW material 1T-TiSe$_2$. Using time- and angle-resolved photoemission spectroscopy (TrARPES) with mid-infrared pumping, we observe a striking pump-induced instantaneous downshift of the valence band maximum (VBM), which is in sharp contrast to the subsequent upward shift on picosecond timescale associated with CDW melting. Most remarkably, the light-induced VBM downshift is observed exclusively in the CDW phase and only when the pump pulse is present, reaching maximum when pumping near resonance with the CDW gap. These observations unequivocally reveal the critical role of CDW in the Floquet engineering of TiSe$_2$. Our work demonstrates how time-periodic drives can synergistically couple to spatially periodic modulations to create non-equilibrium electronic states, establishing a new paradigm for Floquet engineering enabled by spontaneous symmetry breaking.

cond-mat.mtrl-sci↗
arXiv2025

Interaction effects on electronic Floquet spectra: Excitonic effects

Floquet engineering of electronic states by light is a central topic in modern experiments. However, the impact of many-body interactions on the single-electron properties remains unclear in this non-equilibrium situation. We propose that interaction effects could be reasonably understood by performing perturbative expansion in both the pump field and the electron-electron interaction when computing physical quantities. As an example, we apply this approach to semiconductors and show analytically that excitonic effects, i.e., effects of electron-hole interaction, lead to dramatic corrections to the single-electron Floquet spectra even when the excitons are only virtually excited by the pump light. We compute these effects in phosphorene and monolayer MoS$_2$ for time- and angle-resolved photoemission spectroscopy and ultrafast optical experiments.

cond-mat.other↗