arXiv subjects

Qi Luo

Publications and source records attributed to Qi Luo.

At least 37 records · Page 2Linked to original sources

arXiv2025

Towards Global Retrieval Augmented Generation: A Benchmark for Corpus-Level Reasoning

Retrieval-augmented generation (RAG) has emerged as a leading approach to reducing hallucinations in large language models (LLMs). Current RAG evaluation benchmarks primarily focus on what we call local RAG: retrieving relevant chunks from a small subset of documents to answer queries that require only localized understanding within specific text chunks. However, many real-world applications require a fundamentally different capability -- global RAG -- which involves aggregating and analyzing information across entire document collections to derive corpus-level insights (for example, "What are the top 10 most cited papers in 2023?"). In this paper, we introduce GlobalQA -- the first benchmark specifically designed to evaluate global RAG capabilities, covering four core task types: counting, extremum queries, sorting, and top-k extraction. Through systematic evaluation across different models and baselines, we find that existing RAG methods perform poorly on global tasks, with the strongest baseline achieving only 1.51 F1 score. To address these challenges, we propose GlobalRAG, a multi-tool collaborative framework that preserves structural coherence through chunk-level retrieval, incorporates LLM-driven intelligent filters to eliminate noisy documents, and integrates aggregation modules for precise symbolic computation. On the Qwen2.5-14B model, GlobalRAG achieves 6.63 F1 compared to the strongest baseline's 1.51 F1, validating the effectiveness of our method.

cs.CL↗

arXiv2025

Zero-RAG: Towards Retrieval-Augmented Generation with Zero Redundant Knowledge

Retrieval-Augmented Generation has shown remarkable results to address Large Language Models' hallucinations, which usually uses a large external corpus to supplement knowledge to LLMs. However, with the development of LLMs, the internal knowledge of LLMs has expanded significantly, thus causing significant knowledge redundancy between the external corpus and LLMs. On the one hand, the indexing cost of dense retrieval is highly related to the corpus size and thus significant redundant knowledge intensifies the dense retrieval's workload. On the other hand, the redundant knowledge in the external corpus is not helpful to LLMs and our exploratory analysis shows that it instead hurts the RAG performance on those questions which the LLM can answer by itself. To address these issues, we propose Zero-RAG to tackle these challenges. Specifically, we first propose the Mastery-Score metric to identify redundant knowledge in the RAG corpus to prune it. After pruning, answers to "mastered" questions rely primarily on internal knowledge of the LLM. To better harness the internal capacity, we propose Query Router and Noise-Tolerant Tuning to avoid the irrelevant documents' distraction and thus further improve the LLM's utilization of internal knowledge with pruned corpus. Experimental results show that Zero-RAG prunes the Wikipedia corpus by 30\% and accelerates the retrieval stage by 22\%, without compromising RAG's performance.

cs.CL↗

arXiv2025

MARAG-R1: Beyond Single Retriever via Reinforcement-Learned Multi-Tool Agentic Retrieval

Large Language Models (LLMs) excel at reasoning and generation but are inherently limited by static pretraining data, resulting in factual inaccuracies and weak adaptability to new information. Retrieval-Augmented Generation (RAG) addresses this issue by grounding LLMs in external knowledge; However, the effectiveness of RAG critically depends on whether the model can adequately access relevant information. Existing RAG systems rely on a single retriever with fixed top-k selection, restricting access to a narrow and static subset of the corpus. As a result, this single-retriever paradigm has become the primary bottleneck for comprehensive external information acquisition, especially in tasks requiring corpus-level reasoning. To overcome this limitation, we propose MARAG-R1, a reinforcement-learned multi-tool RAG framework that enables LLMs to dynamically coordinate multiple retrieval mechanisms for broader and more precise information access. MARAG-R1 equips the model with four retrieval tools -- semantic search, keyword search, filtering, and aggregation -- and learns both how and when to use them through a two-stage training process: supervised fine-tuning followed by reinforcement learning. This design allows the model to interleave reasoning and retrieval, progressively gathering sufficient evidence for corpus-level synthesis. Experiments on GlobalQA, HotpotQA, and 2WikiMultiHopQA demonstrate that MARAG-R1 substantially outperforms strong baselines and achieves new state-of-the-art results in corpus-level reasoning tasks.

cs.CL↗

arXiv2025

R3-RAG: Learning Step-by-Step Reasoning and Retrieval for LLMs via Reinforcement Learning

Retrieval-Augmented Generation (RAG) integrates external knowledge with Large Language Models (LLMs) to enhance factual correctness and mitigate hallucination. However, dense retrievers often become the bottleneck of RAG systems due to their limited parameters compared to LLMs and their inability to perform step-by-step reasoning. While prompt-based iterative RAG attempts to address these limitations, it is constrained by human-designed workflows. To address these limitations, we propose $\textbf{R3-RAG}$, which uses $\textbf{R}$einforcement learning to make the LLM learn how to $\textbf{R}$eason and $\textbf{R}$etrieve step by step, thus retrieving comprehensive external knowledge and leading to correct answers. R3-RAG is divided into two stages. We first use cold start to make the model learn the manner of iteratively interleaving reasoning and retrieval. Then we use reinforcement learning to further harness its ability to better explore the external retrieval environment. Specifically, we propose two rewards for R3-RAG: 1) answer correctness for outcome reward, which judges whether the trajectory leads to a correct answer; 2) relevance-based document verification for process reward, encouraging the model to retrieve documents that are relevant to the user question, through which we can let the model learn how to iteratively reason and retrieve relevant documents to get the correct answer. Experimental results show that R3-RAG significantly outperforms baselines and can transfer well to different retrievers. We release R3-RAG at https://github.com/Yuan-Li-FNLP/R3-RAG.

cs.CL↗

arXiv2025

Decompile-Bench: Million-Scale Binary-Source Function Pairs for Real-World Binary Decompilation

Recent advances in LLM-based decompilers have been shown effective to convert low-level binaries into human-readable source code. However, there still lacks a comprehensive benchmark that provides large-scale binary-source function pairs, which is critical for advancing the LLM decompilation technology. Creating accurate binary-source mappings incurs severe issues caused by complex compilation settings and widespread function inlining that obscure the correspondence between binaries and their original source code. Previous efforts have either relied on used contest-style benchmarks, synthetic binary-source mappings that diverge significantly from the mappings in real world, or partially matched binaries with only code lines or variable names, compromising the effectiveness of analyzing the binary functionality. To alleviate these issues, we introduce Decompile-Bench, the first open-source dataset comprising two million binary-source function pairs condensed from 100 million collected function pairs, i.e., 450GB of binaries compiled from permissively licensed GitHub projects. For the evaluation purposes, we also developed a benchmark Decompile-Bench-Eval including manually crafted binaries from the well-established HumanEval and MBPP, alongside the compiled GitHub repositories released after 2025 to mitigate data leakage issues. We further explore commonly-used evaluation metrics to provide a thorough assessment of the studied LLM decompilers and find that fine-tuning with Decompile-Bench causes a 20% improvement over previous benchmarks in terms of the re-executability rate. Our code and data has been released in HuggingFace and Github. https://github.com/albertan017/LLM4Decompile

cs.SE↗

arXiv2025

Deterministic and Scalable Coupling of Single 4H-SiC Spin Defects into Bullseye Cavities

Silicon carbide (SiC) has attracted significant attention as a promising quantum material due to its ability to host long-lived, optically addressable color centers with solid-state photonic interfaces. The CMOS compatibility of 4H-SiCOI (silicon-carbide-on-insulator) makes it an ideal platform for integrated quantum photonic devices and circuits. However, the deterministic integration of single spin defects into high-performance photonic cavities on this platform has remained a key challenge. In this work, we demonstrate the deterministic and scalable coupling of both ensemble (PL4) and single PL6 spin defects into monolithic bullseye cavities on the 4H-SiCOI platform. By tuning the cavity resonance, we achieve a 40-fold enhancement of the zero-phonon line (ZPL) intensity from ensemble PL4 defects, corresponding to a Purcell factor of approximately 5.0. For deterministically coupled single PL6 defects, we observe a threefold increase in the saturated photon count rate, confirm single-photon emission, and demonstrate coherent control of the spin state through optically detected magnetic resonance (ODMR), resonant excitation, and Rabi oscillations. These advancements establish a viable pathway for developing scalable, high-performance SiC-based quantum photonic circuits.

quant-ph↗

arXiv2025

Strain Enhanced Spin Readout Contrast in Silicon Carbide Membranes

Quantum defects in solids have emerged as a transformative platform for advancing quantum technologies. A key requirement for these applications is achieving high-fidelity single-spin readout, particularly at room temperature for quantum biosensing. Here, we demonstrate through ab initio simulations of a primary quantum defect in 4H silicon carbide that strain is an effective control parameter for significantly enhancing readout contrast. We validate this principle experimentally by inducing local strain in silicon carbide-on-insulator membranes, achieving a readout contrast exceeding 60% while preserving the favorable coherence properties of single spins. Our findings establish strain engineering as a powerful and versatile strategy for optimizing coherent spin-photon interfaces in solid-state quantum systems.

quant-ph↗

arXiv2025

Evaluation of Switching Technologies for Reflective and Transmissive RISs at Sub-THz Frequencies

For the upcoming 6G wireless networks, reconfigurable intelligent surfaces are an essential technology, enabling dynamic beamforming and signal manipulation in both reflective and transmissive modes. It is expected to utilize frequency bands in the millimeter-wave and THz, which presents unique opportunities but also significant challenges. The selection of switching technologies that can support high-frequency operation with minimal loss and high efficiency is particularly complex. In this work, we demonstrate the potential of advanced components such as Schottky diodes, memristor switches, liquid metal-based switches, phase change materials, and RF-SOI technology in RIS designs as an alternative to overcome limitations inherent in traditional technologies in D-band (110-170 GHz).

cs.ET↗

arXiv2025

Dynamic Order Fulfillment in Last-Mile Drone Delivery under Demand Uncertainty

Drones have attracted growing interest in last-mile delivery due to their potential to significantly reduce costs and enhance operational flexibility, particularly in areas of sparse and uncertain demand where traditional truck delivery proves inefficient. This paper addresses the dynamic order fulfillment problem faced by a retailer operating a fleet of drones to service delivery requests that arrive stochastically. These delivery requests may vary in package profiles, delivery locations, and urgency. We adopt a rolling-horizon framework for order fulfillment and devise a two-stage stochastic program aimed at strategically managing existing orders while considering incoming requests that are subject to various uncertainties. A significant challenge in deploying the envisioned two-stage model lies in its incorporation of vehicle routing constraints, on which exact or brute-force methods are computationally inefficient and unsuitable for real-time operational decisions. To address this, we propose an accelerated L-shaped algorithm that (i) reduces the branching tree size, (ii) replaces exact second-stage solutions with heuristic estimates, and (iii) adapts an alternating strategy for adding optimality cuts. The proposed heuristic demonstrates remarkable performance superiority over the exact method, achieving a 20-fold reduction in average runtime while maintaining an average optimality gap of less than 1\%. We apply the algorithm to a wide range of instances to evaluate the benefits of postponing orders for batch service using the stochastic model. Our results show potential long-term cost savings of up to 20\% when demand uncertainty is explicitly considered in order fulfillment decisions. Meanwhile, the derived savings tend to diminish as the uncertainty increases in order arrivals.

math.OC↗

arXiv2025

Optimal Krylov On Average

We propose an adaptive randomized truncation estimator for Krylov subspace methods that optimizes the trade-off between the solution variance and the computational cost, while remaining unbiased. The estimator solves a constrained optimization problem to compute the truncation probabilities on the fly, with minimal computational overhead. The problem has a closed-form solution when the improvement of the deterministic algorithm satisfies a diminishing returns property. We prove that obtaining the optimal adaptive truncation distribution is impossible in the general case. Without the diminishing return condition, our estimator provides a suboptimal but still unbiased solution. We present experimental results in GP hyperparameter training and competitive physics-informed neural networks problem to demonstrate the effectiveness of our approach.

math.NA↗

arXiv2024

Tunable cavity coupling to spin defects in 4H-silicon-carbide-on-insulator platform

Silicon carbide (SiC) has attracted significant attention as a promising quantum material due to its ability to host long-lived, optically addressable color centers with solid-state photonic interfaces. The CMOS compatibility of 4H-SiCOI (silicon-carbide-on-insulator) makes it an ideal platform for integrated quantum photonic devices and circuits. While micro-ring cavities have been extensively studied in SiC and other materials, the integration of 4H-SiC spin defects into these critical structures, along with continuous mode tunability, remains unexplored. In this work, we demonstrate the integration of PL4 divacancy spin defects into tunable micro-ring cavities in scalable thin-film 4H-SiC nanophotonics. Comparing on- and off-resonance conditions, we observed an enhancement of the Purcell factor by approximately 5.0. This enhancement effectively confined coherent photons within the coupled waveguide, leading to a twofold increase in the ODMR (optically detected magnetic resonance) contrast and coherent control of PL4 spins. These advancements lay the foundation for developing SiC-based quantum photonic circuits.

quant-ph↗

arXiv2024

Room-Temperature entangled quantum processor on integrated semiconductor photonics platform

The rise of the 4H-silicon-carbide-on-insulator (SiCOI) platform marks a promising pathway towards the realization of monolithic quantum photonic networks. However, the challenge of establishing room-temperature entangled registers on these integrated photonics platforms remains unresolved. Herein, we demonstrate the first entangled processor on the SiCOI platform. We show that both deterministic generation of single divacancy electron spins and near-unity spin initialization of a single $^{13}$C nuclear spin can be achieved on SiCOI at room temperature. Besides coherently manipulating the single nuclear spin, a maximally entangled state with a fidelity of 0.89 has been prepared on this CMOS-compatible semiconductor-integrated photonics system. This work establishes the foundation for compact and on-chip solutions within existing defect-based computing and sensing protocols, positioning the SiCOI platform as the most promising candidate for integrated monolithic quantum photonic networks.

quant-ph↗

arXiv2024

LLM4Decompile: Decompiling Binary Code with Large Language Models

Decompilation aims to convert binary code to high-level source code, but traditional tools like Ghidra often produce results that are difficult to read and execute. Motivated by the advancements in Large Language Models (LLMs), we propose LLM4Decompile, the first and largest open-source LLM series (1.3B to 33B) trained to decompile binary code. We optimize the LLM training process and introduce the LLM4Decompile-End models to decompile binary directly. The resulting models significantly outperform GPT-4o and Ghidra on the HumanEval and ExeBench benchmarks by over 100% in terms of re-executability rate. Additionally, we improve the standard refinement approach to fine-tune the LLM4Decompile-Ref models, enabling them to effectively refine the decompiled code from Ghidra and achieve a further 16.2% improvement over the LLM4Decompile-End. LLM4Decompile demonstrates the potential of LLMs to revolutionize binary code decompilation, delivering remarkable improvements in readability and executability while complementing conventional tools for optimal results. Our code, dataset, and models are released at https://github.com/albertan017/LLM4Decompile

cs.PL↗

arXiv2024

Design and Implementation of Ultra-Wideband Dual-Polarized Conformal Phased Array

This letter presents an innovative design and implementation method for an ultra-wideband dual-polarized conformal phased array. The performance of single-layer continuous dipoles under bending conditions is evaluated through characteristic mode analysis to understand the impact of curvature on operational stability. This analysis provides design insights into the maximum allowable curvature while maintaining stable dipole performance. Subsequently, a single-layer metal radiation structure is designed as the array element, and simulations indicate that it supports ultra-wideband operation with dual polarizations. Using this element, an 8 * 8 ultra-wideband dual-polarized cylindrical-conformal array (UDCA) is developed, demonstrating stable performance even at a curvature radius as small as 100 mm. A physical prototype is fabricated cost-effectively using novel manufacturing techniques that stack a three-layer conformal substrate. Experimental result demonstrates that the proposed UDCA with a 1.2λ curvature radius operates at 3.6~9.6 GHz (90.9%) and achieves 60° wide-scanning in E-/H-planes, which provides a practical and promising solution for conformal array applications.

cs.IT↗

arXiv2024

The Peculiar Precursor of a Gamma-Ray Burst from a Binary Merger Involving a Magnetar

The milestone discovery of GW 170817-GRB 170817A-AT 2017gfo has shown that gravitational wave (GW) could be produced during the merger of neutron star-neutron star/black hole and that in electromagnetic (EM) wave a gamma-ray burst (GRB) and a kilonova (KN) are generated in sequence after the merger. Observationally, however, EM property before the merger phase is still unclear. Here we report a peculiar precursor in a KN-associated long-duration GRB 211211A, providing evidence of the EM before the merger. This precursor lasts $\sim$ 0.2 s, and the waiting time between the precursor and the main burst is $\sim$ 1 s, comparable to that between GW 170817 and GRB 170817A. The spectrum of the precursor could be well fit with a non-thermal cutoff power-law model instead of a blackbody. Especially, a $\sim$22 Hz Quasi-Periodic Oscillation candidate ($\sim 3σ$) is detected in the precursor. These temporal and spectral properties indicate that this precursor is probably produced by a catastrophic flare accompanying with magnetoelastic or crustal oscillations of a magnetar in binary compact merger. The strong magnetic field of the magnetar can also account for the prolonged duration of GRB 211211A. However, it poses a challenge to reconcile the rather short lifetime of a magnetar with the rather long spiraling time of a binary neutron star system only by the GW radiation before merger.

astro-ph.HE↗

arXiv2024

SkyGrid: Energy-Flow Optimization at Harmonized Aerial Intersections

The rapid evolution of urban air mobility (UAM) is reshaping the future of transportation by integrating aerial vehicles into urban transit systems. The design of aerial intersections plays a critical role in the phased development of UAM systems to ensure safe and efficient operations in air corridors. This work adapts the concept of rhythmic control of connected and automated vehicles (CAVs) at unsignalized intersections to address complex traffic control problems. This control framework assigns UAM vehicles to different movement groups and significantly reduces the computation of routing strategies to avoid conflicts. In contrast to ground traffic, the objective is to balance three measures: minimizing energy utilization, maximizing intersection flow (throughput), and maintaining safety distances. This optimization method dynamically directs traffic with various demands, considering path assignment distributions and segment-level trajectory coefficients for straight and curved paths as control variables. To the best of our knowledge, this is the first work to consider a multi-objective optimization approach for unsignalized intersection control in the air and to propose such optimization in a rhythmic control setting with time arrival and UAM operational constraints. A sensitivity analysis with respect to inter-platoon safety and straight/left demand balance demonstrates the effectiveness of our method in handling traffic under various scenarios.

eess.SY↗

arXiv2024

Prompt-based Code Completion via Multi-Retrieval Augmented Generation

Automated code completion, aiming at generating subsequent tokens from unfinished code, has been significantly benefited from recent progress in pre-trained Large Language Models (LLMs). However, these models often suffer from coherence issues and hallucinations when dealing with complex code logic or extrapolating beyond their training data. Existing Retrieval Augmented Generation (RAG) techniques partially address these issues by retrieving relevant code with a separate encoding model where the retrieved snippet serves as contextual reference for code completion. However, their retrieval scope is subject to a singular perspective defined by the encoding model, which largely overlooks the complexity and diversity inherent in code semantics. To address this limitation, we propose ProCC, a code completion framework leveraging prompt engineering and the contextual multi-armed bandits algorithm to flexibly incorporate and adapt to multiple perspectives of code. ProCC first employs a prompt-based multi-retriever system which crafts prompt templates to elicit LLM knowledge to understand code semantics with multiple retrieval perspectives. Then, it adopts the adaptive retrieval selection algorithm to incorporate code similarity into the decision-making process to determine the most suitable retrieval perspective for the LLM to complete the code. Experimental results demonstrate that ProCC outperforms state-of-the-art code completion technique by 8.6% on our collected open-source benchmark suite and 10.1% on the private-domain benchmark suite collected from a billion-user e-commerce company in terms of Exact Match. ProCC also allows augmenting fine-tuned techniques in a plug-and-play manner, yielding 5.6% improvement over our studied fine-tuned model.

cs.SE↗

arXiv2024

Adaptive Polynomial Chaos Expansion for Uncertainty Quantification and Optimization of Horn Antennas at SubTHz Frequencies

Sub-terahertz (subTHz) antennas will play an important role in the next generations of wireless communication systems. However, when comes to the subTHz frequency spectrum, the antenna fabrication tolerance needs to be accurately considered during the design stage. The classic approach to studying the average performance of an antenna design considering fabrication tolerances is through the use of the Monte-Carlo (MC) method. In this paper, we propose an adaptive polynomial chaos expansion (PCE) method for the uncertainty quantification analysis of subTHz horn antennas with flat-top radiation patterns. The proposed method builds a surrogate model of the antenna's response to electromagnetic (EM) excitation and estimates its statistical moments with accuracy close to the reference MC method, but with a much smaller computational complexity of roughly two orders of magnitude. Moreover, the surrogate model based on PCE can substitute full-wave EM solvers in producing samples for electromagnetic quantities of interest, resulting in significant computational efficiency gains during optimization tasks. To this end, we successfully combined PCE with the particle swarm optimization method to design the free parameters of a horn antenna at $95$ GHz for a flat-top gain.

cs.IT↗