Searcharxiv⌕ Search

arXiv subjects

Yu Zeng

Publications and source records attributed to Yu Zeng.

At least 37 records · Page 2Linked to original sources
arXiv2026

AppellateGen: A Benchmark for Appellate Legal Judgment Generation

Legal judgment generation is a critical task in legal intelligence. However, existing research in legal judgment generation has predominantly focused on first-instance trials, relying on static fact-to-verdict mappings while neglecting the dialectical nature of appellate (second-instance) review. To address this, we introduce AppellateGen, a benchmark for second-instance legal judgment generation comprising 7,351 case pairs. The task requires models to draft legally binding judgments by reasoning over the initial verdict and evidentiary updates, thereby modeling the causal dependency between trial stages. We further propose a judicial Standard Operating Procedure (SOP)-based Legal Multi-Agent System (SLMAS) to simulate judicial workflows, which decomposes the generation process into discrete stages of issue identification, retrieval, and drafting. Experimental results indicate that while SLMAS improves logical consistency, the complexity of appellate reasoning remains a substantial challenge for current LLMs. The dataset and code are publicly available at: https://anonymous.4open.science/r/AppellateGen-5763.

cs.CY↗
arXiv2026

Vision-DeepResearch: Incentivizing DeepResearch Capability in Multimodal Large Language Models

Multimodal large language models (MLLMs) have achieved remarkable success across a broad range of vision tasks. However, constrained by the capacity of their internal world knowledge, prior work has proposed augmenting MLLMs by ``reasoning-then-tool-call'' for visual and textual search engines to obtain substantial gains on tasks requiring extensive factual information. However, these approaches typically define multimodal search in a naive setting, assuming that a single full-level or entity-level image query and few text query suffices to retrieve the key evidence needed to answer the question, which is unrealistic in real-world scenarios with substantial visual noise. Moreover, they are often limited in the reasoning depth and search breadth, making it difficult to solve complex questions that require aggregating evidence from diverse visual and textual sources. Building on this, we propose Vision-DeepResearch, which proposes one new multimodal deep-research paradigm, i.e., performs multi-turn, multi-entity and multi-scale visual and textual search to robustly hit real-world search engines under heavy noise. Our Vision-DeepResearch supports dozens of reasoning steps and hundreds of engine interactions, while internalizing deep-research capabilities into the MLLM via cold-start supervision and RL training, resulting in a strong end-to-end multimodal deep-research MLLM. It substantially outperforming existing multimodal deep-research MLLMs, and workflows built on strong closed-source foundation model such as GPT-5, Gemini-2.5-pro and Claude-4-Sonnet. The code will be released in https://github.com/Osilly/Vision-DeepResearch.

cs.CV↗
arXiv2026

NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation

Standard diffusion corrupts data using Gaussian noise whose Fourier coefficients have random magnitudes and random phases. While effective for unconditional or text-to-image generation, corrupting phase components destroys spatial structure, making it ill-suited for tasks requiring geometric consistency, such as re-rendering, simulation enhancement, and image-to-image translation. We introduce Phase-Preserving Diffusion (ϕ-PD), a model-agnostic reformulation of the diffusion process that preserves input phase while randomizing magnitude, enabling structure-aligned generation without architectural changes or additional parameters. We further propose Frequency-Selective Structured (FSS) noise, which provides continuous control over structural rigidity via a single frequency-cutoff parameter. ϕ-PD adds no inference-time cost and is compatible with any diffusion model for images or videos. Across photorealistic and stylized re-rendering, as well as sim-to-real enhancement for driving planners, ϕ-PD produces controllable, spatially aligned results. When applied to the CARLA simulator, ϕ-PD significantly improves sim-to-real planner transfer performance. The method is complementary to existing conditioning approaches and broadly applicable to image-to-image and video-to-video generation. Videos, additional examples, and code are available on our \href{https://yuzeng-at-tri.github.io/ppd-page/}{project page}.

cs.CV↗
arXiv2026

Vision-DeepResearch Benchmark: Rethinking Visual and Textual Search for Multimodal Large Language Models

Multimodal Large Language Models (MLLMs) have advanced VQA and now support Vision-DeepResearch systems that use search engines for complex visual-textual fact-finding. However, evaluating these visual and textual search abilities is still difficult, and existing benchmarks have two major limitations. First, existing benchmarks are not visual search-centric: answers that should require visual search are often leaked through cross-textual cues in the text questions or can be inferred from the prior world knowledge in current MLLMs. Second, overly idealized evaluation scenario: On the image-search side, the required information can often be obtained via near-exact matching against the full image, while the text-search side is overly direct and insufficiently challenging. To address these issues, we construct the Vision-DeepResearch benchmark (VDR-Bench) comprising 2,000 VQA instances. All questions are created via a careful, multi-stage curation pipeline and rigorous expert review, designed to assess the behavior of Vision-DeepResearch systems under realistic real-world conditions. Moreover, to address the insufficient visual retrieval capabilities of current MLLMs, we propose a simple multi-round cropped-search workflow. This strategy is shown to effectively improve model performance in realistic visual retrieval scenarios. Overall, our results provide practical guidance for the design of future multimodal deep-research systems. The code will be released in https://github.com/Osilly/Vision-DeepResearch.

cs.CV↗
arXiv2026

Agentic Jigsaw Interaction Learning for Enhancing Visual Perception and Reasoning in Vision-Language Models

Although current large Vision-Language Models (VLMs) have advanced in multimodal understanding and reasoning, their fundamental perceptual and reasoning abilities remain limited. Specifically, even on simple jigsaw tasks, existing VLMs perform near randomly, revealing deficiencies in core perception and reasoning capabilities. While high-quality vision-language data can enhance these capabilities, its scarcity and limited scalability impose significant constraints. To address this, we propose AGILE, an Agentic jiGsaw Interaction Learning for Enhancing visual perception and reasoning in VLMs. AGILE formulates jigsaw solving as an interactive process, enabling the model to progressively engage with the environment. At each step, the model generates executable code to perform an action based on the current state, while the environment provides fine-grained visual feedback to guide task completion. Through this iterative cycle of observation and interaction, the model incrementally improves its perceptual and reasoning capabilities via exploration and feedback. Experimental results show that AGILE not only substantially boosts performance on jigsaw tasks of varying complexity (e.g., increasing accuracy from 9.5% to 82.8% under the 2 $\times$ 2 setting) but also demonstrates strong generalization across 9 general vision tasks, achieving an average improvement of 3.1%. These results indicate notable enhancements in both perceptual and reasoning abilities. This work opens a new avenue for advancing reasoning and generalization in multimodal models and provides an efficient, scalable solution to the scarcity of multimodal reinforcement learning data. The code and datasets is available at https://github.com/yuzeng0-0/AGILE .

cs.AI↗
arXiv2026

Internalizing Meta-Experience into Memory for Guided Reinforcement Learning in Large Language Models

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as an effective approach for enhancing the reasoning capabilities of Large Language Models (LLMs). Despite its efficacy, RLVR faces a meta-learning bottleneck: it lacks mechanisms for error attribution and experience internalization intrinsic to the human learning cycle beyond practice and verification, thereby limiting fine-grained credit assignment and reusable knowledge formation. We term such reusable knowledge representations derived from past errors as meta-experience. Based on this insight, we propose Meta-Experience Learning (MEL), a novel framework that incorporates self-distilled meta-experience into the model's parametric memory. Building upon standard RLVR, we introduce an additional design that leverages the LLM's self-verification capability to conduct contrastive analysis on paired correct and incorrect trajectories, identify the precise bifurcation points where reasoning errors arise, and summarize them into generalizable meta-experience. The meta-experience is further internalized into the LLM's parametric memory by minimizing the negative log-likelihood, which induces a language-modeled reward signal that bridges correct and incorrect reasoning trajectories and facilitates effective knowledge reuse. Experimental results demonstrate that MEL achieves consistent improvements on benchmarks, yielding 3.92%--4.73% Pass@1 gains across varying model sizes.

cs.LG↗
arXiv2026

V2P-Bench: Evaluating Video-Language Understanding with Visual Prompts for Better Human-Model Interaction

Large Vision-Language Models (LVLMs) have made significant strides in the field of video understanding in recent times. Nevertheless, existing video benchmarks predominantly rely on text prompts for evaluation, which often require complex referential language and diminish both the accuracy and efficiency of human model interaction in turn. To address this limitation, we propose V2P-Bench, a robust and comprehensive benchmark for evaluating the ability of LVLMs to understand Video Visual Prompts in human model interaction scenarios. V2P-Bench consists of 980 videos and 1172 well-structured high-quality QA pairs, each paired with manually annotated visual prompt frames. The benchmark spans three main tasks and twelve categories, thereby enabling fine-grained, instance-level evaluation. Through an in-depth analysis of current LVLMs, we identify several key findings: 1) Visual prompts are both more model-friendly and user-friendly in interactive scenarios than text prompts, leading to significantly improved model performance and enhanced user experience. 2) Models are reasonably capable of zero-shot understanding of visual prompts, but struggle with spatiotemporal understanding. Even o1 achieves only 71.8%, far below the human expert score of 88.3%, while most open-source models perform below 60%. 3) LVLMs exhibit pervasive Hack Phenomena in video question answering tasks, which become more pronounced as video length increases and frame sampling density decreases, thereby inflating performance scores artificially. We anticipate that V2P-Bench will not only shed light on these challenges but also serve as a foundational tool for advancing human model interaction and improving the evaluation of video understanding.

cs.CV↗
arXiv2026

Pressure Induced 18 K Superconductivity and Two Superconducting Phases in CuIr2S4

We report pressure-induced superconductivity in the spinel CuIr$_{2}$S$_{4}$ with a transition temperature ($T_{\text{c}}$) reaching \textbf{18.2 K}, establishing a new record for this class of materials and surpassing the decades-old limit of 13.7 K. Our electrical transport and synchrotron X-ray diffraction studies up to 224 GPa reveal the emergence of \textbf{two distinct superconducting phases} from a charge-ordered insulating state. The first phase (SC-I) appears around 18 GPa, and forms a dome-shaped superconducting region in which the resistivity exhibits a pronounced, field- and current-sensitive drop without reaching strict zero above our base temperature. Above 111.8 GPa, a second, lower-$T_{\text{c}}$ phase (SC-II) emerges and coexists with SC-I over a broad pressure range, and SC-II ultimately develops a true zero-resistance state above 122.2 GPa. These superconducting phases are intimately linked to a cascade of structural transitions that systematically distort the frustrated pyrochlore lattice of Ir atoms. Our results expand the potential for superconductivity in spinels and demonstrate a pathway to high-$T_{\text{c}}$ pairing directly from a correlated insulating state driven by lattice tuning.

cond-mat.supr-con↗
arXiv2026

UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision

While Unified Multimodal Models (UMMs) have achieved remarkable success in cross-modal comprehension, a significant gap persists in their ability to leverage such internal knowledge for high-quality generation. We formalize this discrepancy as Conduction Aphasia, a phenomenon where models accurately interpret multimodal inputs but struggle to translate that understanding into faithful and controllable synthesis. To address this, we propose UniCorn, a simple yet elegant self-improvement framework that eliminates the need for external data or teacher supervision. By partitioning a single UMM into three collaborative roles: Proposer, Solver, and Judge, UniCorn generates high-quality interactions via self-play and employs cognitive pattern reconstruction to distill latent understanding into explicit generative signals. To validate the restoration of multimodal coherence, we introduce UniCycle, a cycle-consistency benchmark based on a Text to Image to Text reconstruction loop. Extensive experiments demonstrate that UniCorn achieves comprehensive and substantial improvements over the base model across six general image generation benchmarks. Notably, it achieves SOTA performance on TIIF(73.8), DPG(86.8), CompBench(88.5), and UniCycle while further delivering substantial gains of +5.0 on WISE and +6.5 on OneIG. These results highlight that our method significantly enhances T2I generation while maintaining robust comprehension, demonstrating the scalability of fully self-supervised refinement for unified multimodal intelligence.

cs.CV↗
arXiv2025

Giant-atom quantum acoustodynamics in hybrid superconducting-phononic integrated circuits

We demonstrate a giant atom by coupling a superconducting transmon qubit to a lithium niobate phononic waveguide at two points separated by about 600 acoustic wavelengths, with a propagation delay of 125 ns. The giant atom yields non-Markovian relaxation dynamics characterized by phonon backflow and a frequency-dependent effective decay rate varying four-fold over merely 4 MHz, corresponding to a Purcell factor exceeding 40. Exploiting this frequency-dependent dissipation, we prepare quantum superposition states with high purity. Our results establish phononic integrated circuits as a versatile platform for giant-atom physics, providing highly tunable quantum devices for advanced quantum information processing.

quant-ph↗
arXiv2025

Circuit Quantum Acoustodynamics in a Scalable Phononic Integrated Circuit Architecture

Previous demonstrations of quantum acoustic systems have been limited to isolated devices, with limited capability to route phonons and interconnect multi-port acoustic elements for further extension. Here, we demonstrate a scalable architecture for circuit quantum acoustodynamics (cQAD) by integrating superconducting qubits with suspension-free phononic integrated circuits (PnICs). Coherent coupling between tunable transmon qubits and waveguide-integrated phononic cavities, including Fabry-Perot cavities via monolithic integration and microring cavities via flip-chip assembly, has been achieved, producing a pronounced enhancement of phonon emission with a Purcell factor of ~19. These devices represent elementary building blocks for scalable phononic circuits, establishing the foundation for phonon-based quantum information processors and the testbed for novel quantum acoustic phenomena.

quant-ph↗
arXiv2025

DualVLA: Building a Generalizable Embodied Agent via Partial Decoupling of Reasoning and Action

To build a generalizable Vision-Language-Action (VLA) model with strong reasoning ability, a common strategy is to first train a specialist VLA on robot demonstrations to acquire reliable manipulation skills, and then incorporate mixed annotated robot data together with multimodal data to restore broader reasoning capabilities. However, we observe that the resulting reasoning VLA often suffers from degraded action performance compared to the specialist model before fine-tuning, a phenomenon we refer to as action degeneration. To address this issue, we propose DualVLA, which enhances action performance through carefully designed post-training while still preserving reasoning capability. We first introduce a dual-layer data pruning method that removes redundant embodied reasoning, preventing it from adversely influencing action learning. To further strengthen action generation, we design a dual-teacher adaptive distillation strategy that assigns different supervision signals to different data domains while maintaining reasoning ability. To fill the evaluation gap for generalist VLAs, we also propose VLA Score, which decouples VLA capability into reasoning, intention, action, and alignment dimensions for a more fine-grained assessment. Experiments show that DualVLA achieves an average success rate of 61.0 in SimplerEnv and an average score of 65.4 across eight competitive multimodal benchmarks, demonstrating a stronger balance between precise action execution and multimodal understanding. Project Website: https://costaliya.github.io/DualVLA/.

cs.CV↗
arXiv2025

Acoustically control of integrated optical microrings: from photonic molecule to Mobius strip

Microring resonators (MRRs) are fundamental building blocks of photonic integrated circuits, yet their dynamic reconfiguration has been limited to tuning refractive index or absorption. Here, we demonstrate acoustic control over optical path topology on a lithium niobate on sapphire platform. By launching gigahertz acoustic waves into a hybrid phononic-photonic waveguide, a dynamic Bragg mirror (DBM) is created within the optical path, coupling forward and backward propagating light. Employing a pair of coupled MRRs, we achieve strong coupling between supermodes of the photonic molecule with only milliwatt-level drive power, yielding a cooperativity of 2.46 per milliwatt. At higher power, DBM reflectivity up to 24% is achieved, revealing breakdowns of both the photonic molecule picture and perturbative coupled mode theory, indicating the transformation toward Mobius strip topology. Our work establishes a new dimension for controlling photonic devices, opening pathways toward fully reconfigurable photonic circuits through acoustic drive.

physics.optics↗
arXiv2025

Large-scale programmable phononic integrated circuits

Electronic and photonic chips revolutionized information technology through massive integration of functional elements, yet phonons as fundamental information carriers in solids remain underestimated. Here, we demonstrate large-scale programmable phononic integrated circuits (PnICs) for complex signal processing. We developed a comprehensive library of gigahertz-frequency phononic building blocks that control acoustic wave propagation, polarization, and dispersion. Combining these elements, we demonstrate an ultra-compact 1$\times$128 on-chip acoustic power splitter with unprecedented integration density of 3,000/cm$^2$, a 21-port acoustic frequency demultiplexer with 3.8~MHz resolution, and a four-channel reconfigurable frequency synthesizer. This work establishes scalable phononic integration as the third pillar of information processing alongside electronics and photonics, enabling hybrid chips that combine all three domains for advanced signal processing and quantum information applications.

physics.app-ph↗
arXiv2025

On $p$-parts of conjugacy class sizes and the index of the $p$-core

For a prime $p$ and an arbitrary finite group $G$, we show that if $p^{2}$ does not divide the size of each conjugacy class of \emph{$p$-regular} element (element of order not divisible by $p$) in $G$, then the largest power of $p$ dividing the index $|G:\mathbf{O}_{p}(G)|$ is at most $p^{2}$.

math.GR↗
arXiv2025

Finite groups of non-prime-power order with exactly four character codegrees

For an irreducible complex character \(χ\) of a finite group \(G\), the \emph{codegree} of \(χ\) is defined as the ratio \(|G : \ker(χ)| / χ(1)\), where \(\ker(χ)\) represents the kernel of \(χ\). In this paper, we provide a detailed characterization of finite groups of non-prime power order that have exactly four irreducible character co-degrees.

math.GR↗
arXiv2025

Point-It-Out: Benchmarking Embodied Reasoning for Vision Language Models in Multi-Stage Visual Grounding

Vision-Language Models (VLMs) have demonstrated impressive world knowledge across a wide range of tasks, making them promising candidates for embodied reasoning applications. However, existing benchmarks primarily evaluate the embodied reasoning ability of VLMs through multiple-choice questions based on image annotations -- for example, selecting which trajectory better describes an event in the image. In this work, we introduce the Point-It-Out (PIO) benchmark, a novel benchmark designed to systematically assess the embodied reasoning abilities of VLMs through precise visual grounding. We propose a hierarchical evaluation protocol spanning three stages (S1: referred-object localization, S2: task-driven pointing, and S3: visual trace prediction), with data collected from critical domains for embodied intelligence, including indoor, kitchen, driving, and robotic manipulation scenarios. Extensive experiments with over ten state-of-the-art VLMs reveal several interesting findings. For example, strong general-purpose models such as GPT-4o, while excelling on many benchmarks (e.g., language, perception, and reasoning), underperform compared to some open-source models in precise visual grounding; models such as MoLMO perform well in S1 and S2 but struggle in S3, where requires grounding combined with visual trace planning.

cs.CV↗