arXiv subjects

Rui Li

Publications and source records attributed to Rui Li.

At least 163 records · Page 9Linked to original sources

arXiv2026

FUSAR-KLIP: Towards Multimodal Foundation Models for Remote Sensing

Cross-modal artificial intelligence, represented by visual language models, has achieved significant success in general image understanding. However, a fundamental cognitive inconsistency exists between general visual representation and remote sensing image interpretation: remote sensing images couple topography, terrain, and spatial structure, thereby inherently requiring models to possess deep geoscientific understanding. This cognitive difference is further amplified in synthetic aperture radar (SAR) imagery: while SAR possesses irreplaceable all-weather, all-day observation capabilities, it is constrained by coherent imaging mechanisms, exhibiting significant modal heterogeneity with general images. To address this inconsistency, we propose FUSAR-KLIP, the first knowledge-guided general multimodal foundational model for SAR, along with reusable data and evaluation baselines. Specifically: (1) FUSAR-GEOVL-1M (the first large-scale SAR dataset with complete geographic projection attributes) was constructed, covering multiple satellite platforms, 120,000 images, and 135 cities; (2) Aligned structured text was generated through hierarchical cognitive thought chains, accurately encoding more than 1 million multidimensional semantic information from geomorphological environment and regional attributes to spatial relationships; (3) A self-consistent iterative optimization mechanism was designed to guide cross-modal learning with this knowledge information consistent with human cognition and physical laws in a self-supervised closed loop consisting of contrast, matching, and reconstruction; (4) A unified evaluation benchmark was established in 11 typical downstream tasks in the two major categories of vision and language, and compared with 15 mainstream foundation models.

cs.CV↗

arXiv2026

MIU2Net: weak-lensing mass inversion using deep learning with nested U-structures

One of the primary goals of next-generation gravitational lensing surveys is to measure the large-scale distribution of dark matter, which requires accurate mass inversion to convert weak-lensing shear maps into convergence (kappa) fields. This work develops a mass inversion method tailored for upcoming space missions such as CSST and Euclid, aiming to recover both the mass distribution and the convergence power spectrum with high fidelity. We introduce MIU2Net, a versatile deep-learning framework for kappa-map reconstruction based on the U2-Net architecture. A new loss function is constructed to jointly estimate the convergence field and its frequency-domain energy distribution, effectively balancing optimal mean squared error and optimal power-spectrum recovery. The method incorporates realistic observational effects into shear fields, including shape noise, reduced shear, and complex masks. Under noise levels anticipated for future space-based lensing surveys, MIU2Net recovers the convergence power spectrum with 4% uncertainties up to l approximately 500, significantly outperforming Wiener filtering and MCALens. Beyond two-point statistics, the method accurately reconstructs the convergence distribution, peak centroid, and peak amplitude. Compared to other learning-based approaches such as DeepMass, MIU2Net reduces the root-mean-square error by 5% without smoothing and by 38% with a 1-arcmin smoothing scale. MIU2Net represents a substantial advancement in mass inversion methodology, offering improved accuracy in both RMSE and power-spectrum reconstruction. It provides a promising tool for mapping dark matter environments and large-scale structures in the era of next-generation space lensing surveys.

astro-ph.CO↗

arXiv2026

Integrating Reinforcement Learning with Visual Generative Models: Foundations and Advances

Generative models have made significant progress in synthesizing visual content, including images, videos, and 3D/4D structures. However, they are typically trained with surrogate objectives such as likelihood or reconstruction loss, which often misalign with perceptual quality, semantic accuracy, or physical realism. Reinforcement learning (RL) offers a principled framework for optimizing non-differentiable, preference-driven, and temporally structured objectives. Recent advances demonstrate its effectiveness in enhancing controllability, consistency, and human alignment across generative tasks. This survey provides a systematic overview of RL-based methods for visual content generation. We review the evolution of RL from classical control to its role as a general-purpose optimization tool, and examine its integration into image, video, and 3D/4D generation. Across these domains, RL serves not only as a fine-tuning mechanism but also as a structural component for aligning generation with complex, high-level goals. We conclude with open challenges and future research directions at the intersection of RL and generative modeling.

cs.CV↗

arXiv2026

Be Your Own Red Teamer: Safety Alignment via Self-Play and Reflective Experience Replay

Large Language Models (LLMs) have achieved remarkable capabilities but remain vulnerable to adversarial ``jailbreak'' attacks designed to bypass safety guardrails. Current safety alignment methods depend heavily on static external red teaming, utilizing fixed defense prompts or pre-collected adversarial datasets. This leads to a rigid defense that overfits known patterns and fails to generalize to novel, sophisticated threats. To address this critical limitation, we propose empowering the model to be its own red teamer, capable of achieving autonomous and evolving adversarial attacks. Specifically, we introduce Safety Self- Play (SSP), a system that utilizes a single LLM to act concurrently as both the Attacker (generating jailbreaks) and the Defender (refusing harmful requests) within a unified Reinforcement Learning (RL) loop, dynamically evolving attack strategies to uncover vulnerabilities while simultaneously strengthening defense mechanisms. To ensure the Defender effectively addresses critical safety issues during the self-play, we introduce an advanced Reflective Experience Replay Mechanism, which uses an experience pool accumulated throughout the process. The mechanism employs a Upper Confidence Bound (UCB) sampling strategy to focus on failure cases with low rewards, helping the model learn from past hard mistakes while balancing exploration and exploitation. Extensive experiments demonstrate that our SSP approach autonomously evolves robust defense capabilities, significantly outperforming baselines trained on static adversarial datasets and establishing a new benchmark for proactive safety alignment.

cs.CR↗

arXiv2026

FMVP: Masked Flow Matching for Adversarial Video Purification

Video recognition models remain vulnerable to adversarial attacks, while existing diffusion-based purification methods suffer from inefficient sampling and curved trajectories. Directly regressing clean videos from adversarial inputs often fails to recover faithful content due to the subtle nature of perturbations; this necessitates physically shattering the adversarial structure. Therefore, we propose Flow Matching for Adversarial Video Purification FMVP. FMVP physically shatters global adversarial structures via a masking strategy and reconstructs clean video dynamics using Conditional Flow Matching (CFM) with an inpainting objective. To further decouple semantic content from adversarial noise, we design a Frequency-Gated Loss (FGL) that explicitly suppresses high-frequency adversarial residuals while preserving low-frequency fidelity. We design Attack-Aware and Generalist training paradigms to handle known and unknown threats, respectively. Extensive experiments on UCF-101 and HMDB-51 demonstrate that FMVP outperforms state-of-the-art methods (DiffPure, Defense Patterns (DP), Temporal Shuffling (TS) and FlowPure), achieving robust accuracy exceeding 87% against PGD and 89% against CW attacks. Furthermore, FMVP demonstrates superior robustness against adaptive attacks (DiffHammer) and functions as a zero-shot adversarial detector, attaining AUC-ROC scores of 0.98 for PGD and 0.79 for highly imperceptible CW attacks.

cs.CV↗

arXiv2026

HAPS: Hierarchical LLM Routing with Joint Architecture and Parameter Search

Large language model (LLM) routing aims to exploit the specialized strengths of different LLMs for diverse tasks. However, existing approaches typically focus on selecting LLM architectures while overlooking parameter settings, which are critical for task performance. In this paper, we introduce HAPS, a hierarchical LLM routing framework that jointly searches over model architectures and parameters. Specifically, we use a high-level router to select among candidate LLM architectures, and then search for the optimal parameters for the selected architectures based on a low-level router. We design a parameter generation network to share parameters between the two routers to mutually enhance their capabilities. In the training process, we design a reward-augmented objective to effectively optimize our framework. Experiments on two commonly used benchmarks show that HAPS consistently outperforms strong routing baselines. We have released our code at https://github.com/zihangtian/HAPS.

cs.CL↗

arXiv2026

Efficient Sequential Recommendation for Long Term User Interest Via Personalization

Recent years have witnessed success of sequential modeling, generative recommender, and large language model for recommendation. Though the scaling law has been validated for sequential models, it showed inefficiency in computational capacity when considering real-world applications like recommendation, due to the non-linear(quadratic) increasing nature of the transformer model. To improve the efficiency of the sequential model, we introduced a novel approach to sequential recommendation that leverages personalization techniques to enhance efficiency and performance. Our method compresses long user interaction histories into learnable tokens, which are then combined with recent interactions to generate recommendations. This approach significantly reduces computational costs while maintaining high recommendation accuracy. Our method could be applied to existing transformer based recommendation models, e.g., HSTU and HLLM. Extensive experiments on multiple sequential models demonstrate its versatility and effectiveness. Source code is available at \href{https://github.com/facebookresearch/PerSRec}{https://github.com/facebookresearch/PerSRec}.

cs.IR↗

arXiv2025

Electromagnetically-Induced Transparency Bridges Disconnected Light-Harvesting Networks

The energy-transfer efficiency of the natural photosynthesis system seems to be perfectly optimized during the evolution for millions of years. However, how to enhance the efficiency in the artificial light-harvesting systems is still unclear. In this paper, we investigate the energy-transfer process in the photosystem I (PSI). When there is no effective coupling between the outer antenna (OA) and the reaction center (RC), the two light-harvesting networks are disconnected and thus the energy transfer is inefficient. In order to repair these disconnected networks, we introduce a bridge with three sites between them. We find that by modulating the level structure of the 3-site bridge to be resonant, the energy transfer via the dark state will be enhanced and even outperform the original PSI. Our discoveries may shed light on the designing mechanism of artificial light-harvesting systems.

physics.optics↗

arXiv2025

LabUtopia: High-Fidelity Simulation and Hierarchical Benchmark for Scientific Embodied Agents

Scientific embodied agents play a crucial role in modern laboratories by automating complex experimental workflows. Compared to typical household environments, laboratory settings impose significantly higher demands on perception of physical-chemical transformations and long-horizon planning, making them an ideal testbed for advancing embodied intelligence. However, its development has been long hampered by the lack of suitable simulator and benchmarks. In this paper, we address this gap by introducing LabUtopia, a comprehensive simulation and benchmarking suite designed to facilitate the development of generalizable, reasoning-capable embodied agents in laboratory settings. Specifically, it integrates i) LabSim, a high-fidelity simulator supporting multi-physics and chemically meaningful interactions; ii) LabScene, a scalable procedural generator for diverse scientific scenes; and iii) LabBench, a hierarchical benchmark spanning five levels of complexity from atomic actions to long-horizon mobile manipulation. LabUtopia supports 30 distinct tasks and includes more than 200 scene and instrument assets, enabling large-scale training and principled evaluation in high-complexity environments. We demonstrate that LabUtopia offers a powerful platform for advancing the integration of perception, planning, and control in scientific-purpose agents and provides a rigorous testbed for exploring the practical capabilities and generalization limits of embodied intelligence in future research.

cs.RO↗

arXiv2025

Total and dark mass from observations of galaxy centers with Machine Learning

The galaxy total mass inside the effective radius encode important information on the dark matter and galaxy evolution model. Total "central" masses can be inferred via galaxy dynamics or with gravitational lensing, but these methods have limitations. We propose a novel approach, based on Random Forest, to make predictions on the total and dark matter content of galaxies using simple observables from imaging and spectroscopic surveys. We use catalogs of multi-band photometry, sizes, stellar mass, kinematic "measurements" (features) and dark matter (targets) of simulated galaxies, from Illustris-TNG100 hydrodynamical simulation, to train a Mass Estimate machine Learning Algorithm (Mela). We separate the simulated sample in passive early-type galaxies (ETGs), both "normal" and "dwarf", and active late-type galaxies (LTGs) and show that the mass estimator can accurately predict the galaxy dark masses inside the effective radius in all samples. We finally test the mass estimator against the central mass estimates of a series of low redshift (z$\leq$0.1) datasets, including SPIDER, MaNGA/DynPop and SAMI dwarf galaxies, derived with standard dynamical methods based on Jeans equations. Dynamical masses are reproduced within 0.30 dex ($\sim2σ$), with a limited fraction of outliers and almost no bias. This is independent of the sophistication of the kinematical data collected (fiber vs. 3D spectroscopy) and the dynamical analysis adopted (radial vs. axisymmetric Jeans equations, virial theorem). This makes Mela a powerful alternative to predict the mass of galaxies of massive stage-IV surveys' datasets.

astro-ph.GA↗

arXiv2025

Mechanizing a Proof-Relevant Logical Relation for Timed Message-Passing Protocols

Semantic typing has become a powerful tool for program verification, applying the technique of logical relations as not only a proof method, but also a device for prescribing program behavior. In recent work, Yao et al. scaled semantic typing to the verification of timed message-passing protocols, which are prevalent in, e.g., IoT and real-time systems applications. The appeal of semantic typing in this context is precisely because of its ability to support typed and untyped program components alike -- including physical objects -- which caters to the heterogeneity of these applications. Another demand inherent to these applications is timing: constraining the time or time window within which a message exchange must happen. Yao et al. equipped their logical relation not only with temporal predicates, but also with computable trajectories, to supply the evidence that an inhabitant can step from one time point to another one. While Yao et al. provide the formalization for such a verification tool, it lacks a mechanization. Mechanizing the system would not only provide a machine proof for it, but also facilitate scalability for future extensions and applications. This paper tackles the challenge of mechanizing the resulting proof-relevant logical relation in a proof assistant. allowing trajectories to be interleaved, partitioned, and concatenated, while the intended equality on trajectories is the equality of their graphs when seen as processes indexed by time. Unfortunately, proof assistants based on intensional type theory only have modest support for such equations, forcing a prolific use of transports. This paper reports on the process of mechanizing Yao et al.'s results, comprising the logical relation, the algebra of computable trajectories with supporting lemmas, and the fundamental theorem of the logical relation, in the Rocq theorem prover.

cs.PL↗

arXiv2025

Theoretical study on excited states of ICl+ molecular ion considering spin-orbit coupling

The electronic structure of the ICl+ molecular ion is investigated by using high-level multireference configuration interaction (MRCI) method. To improve computational accuracy, Davidson corrections, spin-orbit coupling (SOC), and core-valence electron correlations effects are incorporated into the calculations. The potential energy curves (PECs) of 21 electronic states associated with the two lowest dissociation limits are obtained. The dipole moments (DMs) of the 21 electronic states of ICl+ are systematically studied, and the variations of DMs of the identical symmetry state in the avoided crossing regions are elucidated by analyzing the dominant electronic configuration. With the help of the calculated SOC matrix element, the interaction between crossing states can be elucidated. Spin-orbit coupling matrix elements involving the several states are calculated. By analyzing potential energy curves of these states and the nearby electronic states, the possible predissociation channels are provided. Based on the computed PECs, the spectroscopic constants of bound states are determined. The comparison of the spectroscopic constants including and excluding SOC effect indicates that the SOC effect has an obvious correction to the spectroscopic properties of low-lying states. Finally, the transition properties between excited states and the ground state are studied. Based on the computed transition dipole moments and Franck-Condon Factors, radiative lifetimes for the low-lying vibrational levels of excited states are evaluated.

physics.atom-ph↗

arXiv2025

Quantum Jump Approach for Photosynthetic Energy Transfer with Chemical Reaction and Fluorescence Loss

Recently, the coherent modified Redfield theory (CMRT) has been widely used to simulate the excitation-energy-transfer (EET) processes in photosynthetic systems. However, the numerical simulation of the CMRT is computationally expensive when dealing with large-scale systems, e.g. photosystem I (PSI) and II (PSII). On the other hand, the chemical reaction and fluorescence loss traditionally treated by the non-Hermitian Hamiltonian approach may result in significantly error in a wide range of parameters. To address these issues, we introduce a quantum jump approach (QJA) based on the CMRT to simulate the evolution of photosynthetic complexes including both the chemical reaction and fluorescence loss. The QJA shows higher accuracy and efficiency in simulating the EET processes. The QJA-CMRT approach may provide a powerful tool to design and optimize artificial photosynthetic systems, which benefits future innovation in the field of energy.

physics.chem-ph↗

arXiv2025

Asymptotic stability of planar entropy wave for 3-d Navier-Stokes equations in Eulerian coordinates

We investigate the large-time asymptotic behavior toward the planar entropy wave for the three-dimensional Navier-Stokes equations in Eulerian coordinates, considering two types of initial perturbations -- with and without the assumption that the integral of the initial perturbation is zero. Generic perturbations generate diffusion waves, and structural conditions fail for multi-dimensional Navier-Stokes equations in Eulerian coordinates. These two aspects have posed significant challenges and left the problem unresolved for years. On one hand, since \cite{LX}, the study of the entropy wave has been based on the left-right structural conditions. Without these structural conditions, the decay rates of lower-order terms become too slow to close the {\it a priori} assumption. On the other hand, the presence of diffusion waves yields problematic error terms in the perturbation system. In this work, we introduce a new transformation to ensure that both left-right structural conditions hold for the perturbation system. Additionally, using the fact that the derivative of the entropy wave maintains a fixed sign, we employ well-designed weighted energy estimates to control the slowly decaying terms. This enables us to establish asymptotic stability and derive the optimal decay rate. Furthermore, we address the case of initial perturbations with the zero mass condition and obtain the optimal decay rate by additionally developing a Poincaré type inequality and a key cancellation.

math.AP↗

arXiv2025

FedP$^2$EFT: Federated Learning to Personalize PEFT for Multilingual LLMs

Federated learning (FL) has enabled the training of multilingual large language models (LLMs) on diverse and decentralized multilingual data, especially on low-resource languages. To improve client-specific performance, personalization via the use of parameter-efficient fine-tuning (PEFT) modules such as LoRA is common. This involves a personalization strategy (PS), such as the design of the PEFT adapter structures (e.g., in which layers to add LoRAs and what ranks) and choice of hyperparameters (e.g., learning rates) for fine-tuning. Instead of manual PS configuration, we propose FedP$^2$EFT, a federated learning-to-personalize method for multilingual LLMs in cross-device FL settings. Unlike most existing PEFT structure selection methods, which are prone to overfitting low-data regimes, FedP$^2$EFT collaboratively learns the optimal personalized PEFT structure for each client via Bayesian sparse rank selection. Evaluations on both simulated and real-world multilingual FL benchmarks demonstrate that FedP$^2$EFT largely outperforms existing personalized fine-tuning methods, while complementing other existing FL methods.

cs.CL↗

arXiv2025

High-fidelity all-microwave CZ gate with partial erasure-error detection via a transmon coupler

Entangling gates between neighboring physical qubits are essential for quantum error correction. Implementing them in an all-microwave manner simplifies signal routing and control apparatus of superconducting quantum processors. We propose and experimentally demonstrate an all-microwave controlled-Z (CZ) gate that achieves high fidelity while suppressing residual ZZ interactions. Our approach utilizes a fixed-frequency transmon coupler and multi-path coupling, thereby sufficiently reducing the net transverse interaction between data transmons to suppress residual ZZ interactions. The controlled phase arises from the dispersive frequency shift of the $|gf\rangle$$\unicode{x2013}$$|eg\rangle$ transition between the coupler and one of the data transmons conditioned on the state of the other data transmon. Driving the transitions at the midpoint of two dispersively shifted resonance frequencies induces state-dependent geometric phases to achieve the CZ gate. Crucially, with this scheme, we can maintain a small net transverse interaction between two data transmons while increasing the coupling between the data and coupler transmons to accelerate the CZ-gate speed. Additionally, we measure the coupler state after the gate to detect a subset of decoherence-induced failures that occur during the gate operation. These events constitute erasure errors with known locations, enabling erasure-aware quantum error-correcting codes to improve future logical qubit performance.

quant-ph↗

arXiv2025

Personalized Interpolation: Achieving Efficient Conversion Estimation with Flexible Optimization Windows

Optimizing conversions is crucial in modern online advertising systems, enabling advertisers to deliver relevant products to users and drive business outcomes. However, accurately predicting conversion events remains challenging due to variable time delays between user interactions (e.g., impressions or clicks) and the actual conversions. These delays vary substantially across advertisers and products, necessitating flexible optimization windows tailored to specific conversion behaviors. To address this, we propose a novel \textit{Personalized Interpolation} method that extends existing models based on fixed conversion windows to support flexible advertiser-specific optimization windows. Our method enables accurate conversion estimation across diverse delay distributions without increasing system complexity. We evaluate the effectiveness of the proposed approach through extensive experiments using a real-world ads conversion model. Our results show that this method achieves both high prediction accuracy and improved efficiency compared to existing solutions. This study demonstrates the potential of our Personalized Interpolation method to improve conversion optimization and support a wider range of advertising strategies in large-scale online advertising systems.

cs.LG↗

arXiv2025

A Survey on Deep Text Hashing: Efficient Semantic Text Retrieval with Binary Representation

With the rapid growth of textual content on the Internet, efficient large-scale semantic text retrieval has garnered increasing attention from both academia and industry. Text hashing, which projects original texts into compact binary hash codes, is a crucial method for this task. By using binary codes, the semantic similarity computation for text pairs is significantly accelerated via fast Hamming distance calculations, and storage costs are greatly reduced. With the advancement of deep learning, deep text hashing has demonstrated significant advantages over traditional, data-independent hashing techniques. By leveraging deep neural networks, these methods can learn compact and semantically rich binary representations directly from data, overcoming the performance limitations of earlier approaches. This survey investigates current deep text hashing methods by categorizing them based on their core components: semantic extraction, hash code quality preservation, and other key technologies. We then present a detailed evaluation schema with results on several popular datasets, followed by a discussion of practical applications and open-source tools for implementation. Finally, we conclude by discussing key challenges and future research directions, including the integration of deep text hashing with large language models to further advance the field. The project for this survey can be accessed at https://github.com/hly1998/DeepTextHashing.

cs.IR↗