arXiv subjects

Yifei Ma

Publications and source records attributed to Yifei Ma.

At least 19 recordsLinked to original sources

arXiv2026

Knowledge-structured and evidence-aware semantic self-distillation for glomerular pathology foundation model

Visual foundation models learn rich pathological morphology but lack explicit concept-level supervision. Vision--language pretraining can provide this supervision but requires costly image--text pairs. Here, we introduce GloSAGE, a semantic self-distillation framework that uses lesion profiles predicted by a pretrained visual teacher to construct structured pathology descriptions. Reliability-guided selection identifies image--text anchors, while Knowledge-Structured Pathology Text Synthesis and Evidence-Guided Negative-pair Construction account for partially observed lesions. Image--text alignment trains projection layers while retaining the frozen visual encoder; an intermediate Semantic Transfer Layer supplies features for downstream tasks. Developed on GloMillion and evaluated across seven glomerulus- and case-level cohorts, GloSAGE supported zero-shot lesion recognition, image--text retrieval, hard-negative discrimination and concept-intervention analyses. It outperformed its source visual model in most lesion-recognition tasks and improved lesion grading, while supporting competitive case-level disease classification, injury grading and baseline renal function estimation. Ablations supported the combined contributions of knowledge-structured text and evidence-guided negatives, and the intermediate transfer layer improved mean lesion-recognition ROC-AUC relative to the complete projection output. A frozen language model conditioned on GloSAGE features generated glomerular descriptions. These findings suggest that teacher-derived lesion profiles can provide annotation-efficient semantic supervision for extending a glomerular visual foundation model.

cs.CV↗

arXiv2026

AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning

Large language models (LLMs) achieve strong reasoning performance, which depends critically on inference-time decisions. Yet these decisions are commonly handled by static, one-size-fits-all policies, limiting adaptation to diverse tasks and reasoning stages. Recent adaptive methods partially address this limitation, but they primarily adapt either decoding stochasticity (how the model explores) or reasoning compute (how long the model reasons) in isolation, leaving their interaction within a single reasoning trajectory unmodeled. To address this challenge, we shift toward a within-trajectory joint control view, and instantiate it in AutoCRAT, a decoder-side controller for frozen backbones. Using only signals available during decoding, AutoCRAT jointly adjusts sampling stochasticity and reasoning budget during generation. AutoCRAT operates over a discrete action space and updates control decisions only at semantic boundaries, improving stability while remaining responsive to the evolving reasoning process. Comprehensive evaluation across 6 benchmarks demonstrates that AutoCRAT (I) uses 13.8-52.7% fewer inference tokens on average than recommended static configurations, (II) surpasses recommended static and adaptive baselines by 1.5-4.5% in relative accuracy, and (III) enjoys strong cross-backbone transferability.

cs.AI↗

arXiv2026

Defects encode high-dimensional topological information

In polarization fields, Stokes skyrmions are continuous vectorial textures that encode integer-valued topological invariants across real space, enabling robust optical information encoding under complex perturbations. This topological resilience, however, fails when singular points occur where the Stokes vector has no unique limiting value, placing a fundamental constraint on skyrmion-based information manipulation. Here, we show, paradoxically, that the very defects that destroy conventional resilience can become the carriers of topological information. We introduce the resulting structures as Stokes defect skyrmions, in which singular Stokes responses constitute measurable topological degrees of freedom with theoretically minimal size. We design and realize one class of them using all-dielectric metasurfaces that combine arbitrarily controlled distinguished fast-axis singularities with customized retardance profiles. The resulting fields are then described by high-dimensional integer-valued topological tuples, providing theoretically unbounded information capacity at the nanoscale. As a proof-of-concept demonstration, selected tuple components are mapped to represent predefined alphabetic symbols, realizing controlled high-dimensional information representation within a single optical field. Our results establish Stokes defects as functional units for higher-dimensional topological encoding, expanding the role of defects from failure points to engineerable carriers of optical information.

physics.optics↗

arXiv2026

OPERA: Online Data Pruning for Efficient Retrieval Model Adaptation

Domain-specific finetuning is essential for dense retrievers, yet not all data pairs contribute equally to the learning process. We introduce OPERA, a data pruning framework that exploits this heterogeneity to improve both the effectiveness and efficiency of retrieval model adaptation. We first investigate static pruning (SP), which retains only high-similarity query-document pairs, revealing an intrinsic quality-coverage tradeoff: ranking (NDCG) improves while retrieval (Recall) can degrade due to reduced query diversity. To resolve this tradeoff, we propose a two-stage dynamic pruning (DP) strategy that adaptively modulates sampling probabilities at both query and document levels throughout training, prioritizing high-quality examples while maintaining access to the full training set. Evaluations across eight datasets spanning six domains demonstrate the effectiveness of both approaches: SP improves ranking over standard finetuning (NDCG@10 +0.2 points), while DP achieves the strongest performance on both ranking (NDCG@10 +1.0 points) and retrieval (Recall@20 +0.4 points), with an average rank of 1.38 across all methods. These findings scale to Qwen3-Embedding, an LLM-based dense retriever, confirming architecture-agnostic benefits. Notably, DP reaches comparable performance in less than 50\% of the training time required by standard finetuning.

cs.IR↗

arXiv2026

Integrated photonic computing: towards high-dimensional information processing

The rapid growth of artificial intelligence, coupled with the slowing of Moore's law, is straining computing infrastructure, as CMOS electronics face inherent limits in bandwidth, energy efficiency, and parallelism. Integrated photonic computing encodes and processes information using the phase, amplitude, spatial modes, wavelength channels, and polarisation of guided optical fields, offering a scalable and energy-efficient route beyond charge-based signalling. Here, we review on-chip photonic computing, emphasising the progression from low-dimensional to high-dimensional architectures. At the foundational level, low-dimensional approaches manipulate the phase and amplitude of guided light through Mach-Zehnder interferometers, diffractive structures, microring resonators, and absorptive elements, forming a programmable basis for optical matrix-vector multiplication. Crucially, high-dimensional architectures exploit spatial modes and wavelength channels to carry multiple independent data streams through a single waveguide, achieving higher throughput with moderate hardware overhead. Practical deployment, however, demands more than device innovation. We examine how system-level techniques, from time-wavelength interleaving to hardware-aware training, address energy efficiency, precision, and algorithm-hardware co-design. Five challenges nevertheless remain: electro-optic conversion efficiency, computing parallelism, spatial integration, reconfigurability, and robustness. We highlight emerging topological structures, such as optical skyrmions, as a promising route to fault-tolerant, topologically protected encoding that exploits the largely untapped polarisation degree of freedom. We argue that, by embracing the higher dimensionality of light, photonic computing can offer not merely an incremental improvement but a new paradigm for high-performance, energy-efficient information processing.

physics.optics↗

arXiv2026

Wanderland: Geometrically Grounded Simulation for Open-World Embodied AI

Reproducible closed-loop evaluation remains a major bottleneck in Embodied AI such as visual navigation. A promising path forward is high-fidelity simulation that combines photorealistic sensor rendering with geometrically grounded interaction in complex, open-world urban environments. Although recent video-3DGS methods ease open-world scene capturing, they are still unsuitable for benchmarking due to large visual and geometric sim-to-real gaps. To address these challenges, we introduce Wanderland, a real-to-sim framework that features multi-sensor capture, reliable reconstruction, accurate geometry, and robust view synthesis. Using this pipeline, we curate a diverse dataset of indoor-outdoor urban scenes and systematically demonstrate how image-only pipelines scale poorly, how geometry quality impacts novel view synthesis, and how all of these adversely affect navigation policy learning and evaluation reliability. Beyond serving as a trusted testbed for embodied navigation, Wanderland's rich raw sensor data further allows benchmarking of 3D reconstruction and novel view synthesis models. Our work establishes a new foundation for reproducible research in open-world embodied AI. Project website is at https://ai4ce.github.io/wanderland/.

cs.CV↗

arXiv2026

Skyrmions based on optical anisotropy for topological encoding

The observation of skyrmions across diverse physical domains suggests that they are universal features of S$^{2}$-valued fields, reflecting the ubiquity of topology in the study of the natural world. In this paper, we develop an abstract technique of parameter space dimensionality reduction that extends the skyrmion framework to fields taking values in manifolds of dimension greater than 2, thereby broadening the range of systems that can support skyrmions. To prove that this is more than just a mathematical abstraction, we apply our technique to light-matter interactions, directly encoding skyrmionic structures into the optical anisotropy of spatially varying structured matter by selecting a distinguished axis, an approach fundamentally different from the more commonly known skyrmions formed by director fields in liquid crystals. We experimentally realize such skyrmions using a liquid-crystal-based tunable elliptical retarder array as a proof-of-concept platform and demonstrate complex, reconfigurable skyrmionic states exhibiting topological robustness under artificially introduced stochastic perturbations. Exploiting this robustness, we demonstrate a promising application of skyrmions in topologically protected information storage and provide an easily verifiable rule, which we term the 60° rule, that serves as a practical engineering criterion for guaranteeing robustness against noise and measurement errors.

physics.optics↗

arXiv2026

GloPath: An Entity-Centric Foundation Model for Glomerular Lesion Assessment and Clinicopathological Insights

Glomerular pathology is central to the diagnosis and prognosis of renal diseases, yet the heterogeneity of glomerular morphology and fine-grained lesion patterns remain challenging for current AI approaches. We present GloPath, an entity-centric foundation model trained on over one million glomeruli extracted from 14,049 renal biopsy specimens using multi-scale and multi-view self-supervised learning. GloPath addresses two major challenges in nephropathology: glomerular lesion assessment and clinicopathological insights discovery. For lesion assessment, GloPath was benchmarked across three independent cohorts on 52 tasks, including lesion recognition, grading, few-shot classification, and cross-modality diagnosis-outperforming state-of-the-art methods in 42 tasks (80.8%). In the large-scale real-world study, it achieved an ROC-AUC of 91.51% for lesion recognition, demonstrating strong robustness in routine clinical settings. For clinicopathological insights, GloPath systematically revealed statistically significant associations between glomerular morphological parameters and clinical indicators across 224 morphology-clinical variable pairs, demonstrating its capacity to connect tissue-level pathology with patient-level outcomes. Together, these results position GloPath as a scalable and interpretable platform for glomerular lesion assessment and clinicopathological discovery, representing a step toward clinically translatable AI in renal pathology.

cs.CV↗

arXiv2026

Optimal Design for Human Preference Elicitation

Learning of preference models from human feedback has been central to recent advances in artificial intelligence. Motivated by the cost of obtaining high-quality human annotations, we study efficient human preference elicitation for learning preference models. The key idea in our work is to generalize optimal designs, an approach to computing optimal information-gathering policies, to lists of items that represent potential questions with answers. The policy is a distribution over the lists and we elicit preferences from them proportionally to their probabilities. To show the generality of our ideas, we study both absolute and ranking feedback models on items in the list. We design efficient algorithms for both and analyze them. Finally, we demonstrate that our algorithms are practical by evaluating them on existing question-answering problems.

cs.LG↗

arXiv2026

Topologically robust programmable logic arrays using light and matter skyrmions

Photonic computing offers a low-power, high-bandwidth paradigm for information processing; however, the analogue nature of conventional architectures means that intrinsic noise and fabrication imperfections greatly impact performance, thereby severely limiting scalability. Recent work on optical skyrmions offers a route to overcoming these limitations by exploiting perturbation-resilient topological invariants assigned to the optical field for computation. Crucially, owing to its relative novelty, an architectural perspective on integrating individual components that manipulate topological charge into a functional system remains an important open goal. In this paper, we take concrete steps toward system-level design by introducing a platform-independent architecture for skyrmion-based logic, built around a modular library of topologically robust optical primitives, including generators, converters, registers, and adders. This framework enables the synthesis and arithmetic manipulation of topological numbers within a unified programmable architecture. We then experimentally validate this approach using multichannel arrays, demonstrating accurate charge readout and high robustness against alignment errors and environmental noise. These results provide a scalable foundation for topologically robust programmable logic arrays, paving the way for compact and integrated photonic processing circuits.

physics.optics↗

arXiv2026

General in situ feedback control of cascaded liquid crystal spatial light modulators for structured field generation

Cascaded liquid crystal spatial light modulators provide a versatile strategy for the generation of structured light and matter fields, with applications including optical communications, photonic computing, and topological field engineering. However, experimental imperfections, such as temperature-dependent liquid crystal response, variations between individual pixels, and alignment errors, present significant engineering challenges in generating high-quality fields. Moreover, changes in experimental conditions over time mean that calibrating each component once is insufficient for maintaining long-term, high-quality field generation. To address this, we present a general engineering approach based on a bespoke, physically informed, and manifold-constrained gradient-descent scheme that enables in situ feedback control, compensating for such errors in real time without the need to alter the experimental setup. We further demonstrate the correction efficacy of our proposed strategy through experiments in both spatially varying light and matter field generation, including scenarios in which complex vectorial aberrations are artificially introduced into the setup. Together, these demonstrations underscore the practicality of our method and its suitability for deployment in real-world experimental environments, paving the way for robust operation of cascaded architectures for structured field generation.

physics.optics↗

arXiv2026

Resolving topological obstructions to vectorial structured field control

The use of structured matter, such as optical retarders, for vectorial control is a well-established and widely employed technique in modern optics, and has driven continued advances in the manipulation of complex, spatially varying vectorial fields. However, achieving arbitrary field conversion typically requires the use of cascaded elements, as intrinsic physical and fabrication constraints fundamentally limit individual devices to a restricted subset of transformations. This results in an overall continuous transformation potentially failing to be continuous at the level of the parameters of the cascade, leading to detrimental engineering consequences such as the introduction of complex, discontinuous aberrations that disrupt important topological properties of the underlying matter field. In this work, we establish a novel mathematical framework for analyzing the topological difficulties that emerge in the decomposition of an overall transformation into individual layers, and for determining the minimal depth required to overcome them. The strategy introduced provides a general pathway for optimizing designs for vectorial field control and matter field generation, with particular significance for the manipulation of topological phases in optical polarization fields, such as Stokes skyrmions, where continuity is of vital importance.

physics.optics↗

arXiv2025

Multiplexed vector beam conversion via complex structured matter

Structured light, in which the amplitude, phase, and polarization of an optical field are deliberately tailored in space and time, has enabled unprecedented control over optical fields, paving the way for diverse applications across photonics and optical engineering. However, the prevailing design philosophy, which predominantly focuses on converting a single fixed input into a single desired output, relies on tunability to achieve time-division multiplexing rather than intrinsic design, and is fundamentally incompatible with wavelength-division multiplexing. Here, we propose a general framework for designing structured matter capable of achieving multiple input-output relations simultaneously, thereby enabling passive devices to realize both time-division and wavelength-division multiplexing. Using Stokes skyrmions, which have recently gained attention for their topological properties and promising applications in modern optical communication and computing, as an example, we demonstrate that a simple retarder-diattenuator-retarder cascade can be designed to simultaneously satisfy three arbitrary input-output relations, enabling diverse functionalities within a single passive element. This approach enables complex and multiplexed manipulation of topological numbers, paving the way for high-dimensional on-chip photonic computing based on optical skyrmions.

physics.optics↗

arXiv2025

Speech Audio Generation from dynamic MRI via a Knowledge Enhanced Conditional Variational Autoencoder

Dynamic Magnetic Resonance Imaging (MRI) of the vocal tract has become an increasingly adopted imaging modality for speech motor studies. Beyond image signals, systematic data loss, noise pollution, and audio file corruption can occur due to the unpredictability of the MRI acquisition environment. In such cases, generating audio from images is critical for data recovery in both clinical and research applications. However, this remains challenging due to hardware constraints, acoustic interference, and data corruption. Existing solutions, such as denoising and multi-stage synthesis methods, face limitations in audio fidelity and generalizability. To address these challenges, we propose a Knowledge Enhanced Conditional Variational Autoencoder (KE-CVAE), a novel two-step "knowledge enhancement + variational inference" framework for generating speech audio signals from cine dynamic MRI sequences. This approach introduces two key innovations: (1) integration of unlabeled MRI data for knowledge enhancement, and (2) a variational inference architecture to improve generative modeling capacity. To the best of our knowledge, this is one of the first attempts at synthesizing speech audio directly from dynamic MRI video sequences. The proposed method was trained and evaluated on an open-source dynamic vocal tract MRI dataset recorded during speech. Experimental results demonstrate its effectiveness in generating natural speech waveforms while addressing MRI-specific acoustic challenges, outperforming conventional deep learning-based synthesis approaches.

cs.SD↗

arXiv2025

Full vectorial field sensing using liquid crystal droplet arrays

Determining the amplitude, phase, and polarization profile of light is essential for both fundamental scientific discovery and applications spanning optical metrology, microscopy, astronomy, and optical communication/computing technologies. However, most modern measurement approaches are unable to retrieve such parameters readily, often relying on bulky and expensive hardware, or lacking the capability for single-shot sensing. Here, we introduce a low cost, compact, full vectorial field sensor based on an inkjet-printed nematic liquid crystal droplet array that enables simultaneous measurement of these important characteristics of light. Polarization and intensity are measured via division-of-wavefront polarimetry, exploiting the droplets' spatially varying birefringence, while the phase is reconstructed by treating each droplet as a separate microlens in a Shack-Hartmann-like wavefront sensor configuration. To demonstrate the system's performance, we characterize aberrated dual-wavelength beams carrying distinct intensity, phase, and polarization information, confirming accurate retrieval of the optical field profiles for both spectral components.

physics.optics↗

arXiv2025

DRAFT-RL: Multi-Agent Chain-of-Draft Reasoning for Reinforcement Learning-Enhanced LLMs

Large Language Models (LLMs) have shown impressive capabilities in multi-step reasoning and problem-solving.Recent works introduce multi-agent reflection frameworks where multiple LLM agents critique and refine each other's outputs using reinforcement learning (RL). However, these approaches often rely on single-shot responses and lack structural diversity in reasoning exploration. In this paper, we propose DRAFT-RL, a novel framework that integrates Chain-of-Draft (CoD) reasoning into multi-agent RL training. Instead of generating single responses, each agent produces multiple drafts per query, which are then evaluated by peer agents and a learned reward model to identify the most promising trajectory. These selected drafts are used to refine future reasoning strategies through actor-critic learning.DRAFT-RL enables explicit multi-path exploration, peer-guided reflection, and reward-aligned selection, resulting in more robust and interpretable LLM agent behavior. We evaluate our method on complex reasoning tasks including code synthesis, symbolic math, and knowledge-intensive QA,demonstrating that DRAFT-RL outperforms existing reflective and RL-based agents by significant margins in both accuracy and convergence speed

cs.AI↗

arXiv2025

Perturbation-resilient integer arithmetic using optical skyrmions

The decline of Moore's law coupled with the rise of artificial intelligence has recently motivated research into photonic computing as a high-bandwidth, low-power strategy to accelerate digital electronics. However, many modern-day photonic computing strategies are analog, making them susceptible to noise and intrinsically difficult to scale. Optical skyrmions offer a route to overcoming these limitations through digitization in the form of a discrete topological number that can be assigned to the analog optical field. Apart from an intrinsic robustness against perturbations, optical skyrmions represent a new medium that has yet to be fully exploited for photonic computing, namely spatially varying polarization. Here, we propose and experimentally demonstrate a method for performing perturbation-resilient integer arithmetic with optical skyrmions and passive optical components. To the best of our knowledge, this is the first time such discrete mathematical operations have been directly achieved using optical skyrmions without external energy input.

physics.optics↗

arXiv2025

Inkjet Printed Liquid Crystal Droplet for Complex Beam Manipulation

The inkjet-fabricated liquid crystal (LC) droplet device not only capitalizes on the intrinsic birefringence properties of liquid crystals but also leverages the hemispherical shape of droplet devices on substrates. This configuration facilitates self-alignment of the LC director under the influence of surface tension. The LC droplet devices we fabricated are capable of intricate beam manipulation, encompassing both generation and analysis of light beams. Such devices possess substantial prospective applications in the fields of optical communications and light beam characterization, highlighting their significant potential for advancement in optical technologies.

physics.optics↗