Searcharxiv⌕ Search

arXiv subjects

Li Li

Publications and source records attributed to Li Li.

At least 145 records · Page 8Linked to original sources
arXiv2025

Token Sugar: Making Source Code Sweeter for LLMs through Token-Efficient Shorthand

Large language models (LLMs) have shown exceptional performance in code generation and understanding tasks, yet their high computational costs hinder broader adoption. One important factor is the inherent verbosity of programming languages, such as unnecessary formatting elements and lengthy boilerplate code. This leads to inflated token counts in both input and generated outputs, which increases inference costs and slows down the generation process. Prior work improves this through simplifying programming language grammar, reducing token usage across both code understanding and generation tasks. However, it is confined to syntactic transformations, leaving significant opportunities for token reduction unrealized at the semantic level. In this work, we propose Token Sugar, a concept that replaces frequent and verbose code patterns with reversible, token-efficient shorthand in the source code. To realize this concept in practice, we designed a systematic solution that mines high-frequency, token-heavy patterns from a code corpus, maps each to a unique shorthand, and integrates them into LLM pretraining via code transformation. With this solution, we obtain 799 (code pattern, shorthand) pairs, which can reduce up to 15.1% token count in the source code and is complementary to existing syntax-focused methods. We further trained three widely used LLMs on Token Sugar-augmented data. Experimental results show that these models not only achieve significant token savings (up to 11.2% reduction) during generation but also maintain near-identical Pass@1 scores compared to baselines trained on unprocessed code.

cs.SE↗
arXiv2025

Chain-of-Query: Unleashing the Power of LLMs in SQL-Aided Table Understanding via Multi-Agent Collaboration

Table understanding requires structured, multi-step reasoning. Large Language Models (LLMs) struggle with it due to the structural complexity of tabular data. Recently, multi-agent frameworks for SQL generation have shown promise in tackling the challenges of understanding tabular data, but existing approaches often suffer from limitations such as the inability to comprehend table structure for reliable SQL generation, error propagation that results in invalid queries, and over-reliance on execution correctness. To address these issues, we propose Chain-of-Query (CoQ), a novel multi-agent framework for SQL-aided table understanding. CoQ adopts natural-language-style representations of table schemas to abstract away structural noise and enhance understanding. It employs a clause-by-clause SQL generation strategy to improve query quality and introduces a hybrid reasoning division that separates SQL-based mechanical reasoning from LLM-based logical inference, thereby reducing reliance on execution outcomes. Extensive experiments across four models and five widely used benchmarks demonstrate that CoQ achieves substantial accuracy improvements and significantly lowers invalid SQL rates compared to prior generic LLM-based, SQL-aided, and hybrid baselines, confirming its superior effectiveness in table understanding. The code is available at https://github.com/SongyuanSui/ChainofQuery.

cs.CL↗
arXiv2025

Less is More: Resource-Efficient Low-Rank Adaptation

Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method for Large Language Models (LLMs), but it still incurs notable overhead and suffers from parameter interference in complex datasets. While recent works decouple LoRA update matrices to exploit matrix-wise asymmetry, training costs remain high. We revisit LoRA from the perspective of inter-matrix and intra-layer parameter redundancy and propose Resource-Efficient Low-Rank Adaptation, EffiLoRA, a lightweight and generalizable approach for language, multimodal, and diffusion models. EffiLoRA employs a unified A matrix across all transformer layers and introduces a runtime selective B matrices update to dynamically trade-off the system resource budget and model performance. EffiLoRA consistently outperforms LoRA across diverse modalities, including commonsense reasoning, visual instruction tuning, and image generation, demonstrating improved efficiency and robustness.

cs.CL↗
arXiv2025

Elastic Mixture of Rank-Wise Experts for Knowledge Reuse in Federated Fine-Tuning

Federated fine-tuning offers a promising solution for adapting Large Language Models (LLMs) to downstream tasks while safeguarding data privacy. However, its high computational and communication demands hinder its deployment on resource-constrained devices. In this paper, we propose SmartFed, a resource-efficient federated fine-tuning framework. SmartFed intelligently reuses knowledge embedded in existing LoRA modules, eliminating the need for expensive training from scratch when adapting LLMs to new tasks. To effectively exploit this knowledge and ensure scalability, we introduce the Mixture of Rank-Wise Experts (MoRE). MoRE decomposes LoRA modules into fine-grained rank-level experts. These experts are selectively activated and combined based on input semantics and resource budgets. Moreover, to optimize resource utilization, we present the Elastic Expert Quota Allocation (EEQA). EEQA adaptively allocates expert capacity across parameter matrices based on their contribution to model performance, focusing computing resources on the critical experts. Extensive evaluations across multiple benchmarks demonstrate that SmartFed significantly outperforms existing methods in model performance and training efficiency.

cs.DC↗
arXiv2025

AutoTailor: Automatic and Efficient Adaptive Model Deployment for Diverse Edge Devices

On-device machine learning (ML) has become a fundamental component of emerging mobile applications. Adaptive model deployment delivers efficient inference for heterogeneous device capabilities and performance requirements through customizing neural architectures. SuperNet-based approaches offer a promising solution by generating a large number of model variants from a pre-trained ML model. However, applying SuperNet in existing frameworks suffers from tedious model-aware development and time-consuming hardware-aware profiling, which limits their practical adoption. We present AutoTailor, the first framework to enable automated, end-to-end SuperNet-based adaptive model deployment for edge devices. Unlike manual SuperNet construction, AutoTailor employs a computation graph-guided compilation approach to automatically transform user-provided ML models into SuperNets. To support efficient specialization, AutoTailor incorporates learning-free latency and accuracy predictors, enabling low-cost yet accurate performance prediction. Our extended evaluations demonstrate that AutoTailor reduces the lines of code for SuperNet construction by 11--27$\times$, decreases hardware-aware profiling costs by at least 11$\times$, and achieves up to 15.60\% absolute accuracy improvement and 60.03\% latency reduction compared to state-of-the-art approaches across diverse models and devices.

cs.LG↗
arXiv2025

AuthenLoRA: Entangling Stylization with Imperceptible Watermarks for Copyright-Secure LoRA Adapters

Low-Rank Adaptation (LoRA) offers an efficient paradigm for customizing diffusion models, but its ease of redistribution raises concerns over unauthorized use and the generation of untraceable content. Existing watermarking techniques either target base models or verify LoRA modules themselves, yet they fail to propagate watermarks to generated images, leaving a critical gap in traceability. Moreover, traceability watermarking designed for base models is not tightly coupled with stylization and often introduces visual degradation or high false-positive detection rates. To address these limitations, we propose AuthenLoRA, a unified watermarking framework that embeds imperceptible, traceable watermarks directly into the LoRA training process while preserving stylization quality. AuthenLoRA employs a dual-objective optimization strategy that jointly learns the target style distribution and the watermark-induced distribution shift, ensuring that any image generated with the watermarked LoRA reliably carries the watermark. We further design an expanded LoRA architecture for enhanced multi-scale adaptation and introduce a zero-message regularization mechanism that substantially reduces false positives during watermark verification. Extensive experiments demonstrate that AuthenLoRA achieves high-fidelity stylization, robust watermark propagation, and significantly lower false-positive rates compared with existing approaches. Open-source implementation is available at: https://github.com/ShiFangming0823/AuthenLoRA

cs.CR↗
arXiv2025

COLI: A Hierarchical Efficient Compressor for Large Images

The escalating adoption of high-resolution, large-field-of-view imagery amplifies the need for efficient compression methodologies. Conventional techniques frequently fail to preserve critical image details, while data-driven approaches exhibit limited generalizability. Implicit Neural Representations (INRs) present a promising alternative by learning continuous mappings from spatial coordinates to pixel intensities for individual images, thereby storing network weights rather than raw pixels and avoiding the generalization problem. However, INR-based compression of large images faces challenges including slow compression speed and suboptimal compression ratios. To address these limitations, we introduce COLI (Compressor for Large Images), a novel framework leveraging Neural Representations for Videos (NeRV). First, recognizing that INR-based compression constitutes a training process, we accelerate its convergence through a pretraining-finetuning paradigm, mixed-precision training, and reformulation of the sequential loss into a parallelizable objective. Second, capitalizing on INRs' transformation of image storage constraints into weight storage, we implement Hyper-Compression, a novel post-training technique to substantially enhance compression ratios while maintaining minimal output distortion. Evaluations across two medical imaging datasets demonstrate that COLI consistently achieves competitive or superior PSNR and SSIM metrics at significantly reduced bits per pixel (bpp), while accelerating NeRV training by up to 4 times.

cs.CV↗
arXiv2025

TimeFlow: Towards Stochastic-Aware and Efficient Time Series Generation via Flow Matching Modeling

Generating high-quality time series data has emerged as a critical research topic due to its broad utility in supporting downstream time series mining tasks. A major challenge lies in modeling the intrinsic stochasticity of temporal dynamics, as real-world sequences often exhibit random fluctuations and localized variations. While diffusion models have achieved remarkable success, their generation process is computationally inefficient, often requiring hundreds to thousands of expensive function evaluations per sample. Flow matching has emerged as a more efficient paradigm, yet its conventional ordinary differential equation (ODE)-based formulation fails to explicitly capture stochasticity, thereby limiting the fidelity of generated sequences. By contrast, stochastic differential equation (SDE) are naturally suited for modeling randomness and uncertainty. Motivated by these insights, we propose TimeFlow, a novel SDE-based flow matching framework that integrates a encoder-only architecture. Specifically, we design a component-wise decomposed velocity field to capture the multi-faceted structure of time series and augment the vanilla flow-matching optimization with an additional stochastic term to enhance representational expressiveness. TimeFlow is flexible and general, supporting both unconditional and conditional generation tasks within a unified framework. Extensive experiments across diverse datasets demonstrate that our model consistently outperforms strong baselines in generation quality, diversity, and efficiency.

cs.AI↗
arXiv2025

Gravitational wave cosmology

Gravitational waves (GWs) originating from cosmological sources offer direct insights into the physics of the primordial Universe, the fundamental nature of gravity, and the cosmic expansion of the Universe. In this review paper, we present a comprehensive overview of our recent advances in GW cosmology, supported by the national key research and development program of China, focusing on cosmological GW sources and their implications for fundamental physics and cosmology. We first discuss the generation mechanisms and characteristics of stochastic gravitational wave backgrounds generated by physical processes occurred in the early Universe, including those from inflation, phase transitions, and topological defects, and summarize current and possible future constraints from pulsar timing array and space-based detectors. Next, we explore the formation and observational prospects of primordial black holes as GW sources and their potential connection to dark matter. We then analyze how GWs are affected by large-scale structure, cosmological perturbations, and possible modifications of gravity on GW propagation, and how these effects can be used to test fundamental symmetry of gravity. Finally, we discuss the application of GW standard sirens in measuring the Hubble constant, the expansion history, and dark energy parameters, including their combination with electromagnetic observations. These topics together show how GW observations, especially with upcoming space-based detectors, such as LISA, Taiji, and Tianqin, can provide new information about the physics of the early Universe, cosmological evolution, and the nature of gravity.

gr-qc↗
arXiv2025

LLM-Powered Text-Attributed Graph Anomaly Detection via Retrieval-Augmented Reasoning

Anomaly detection on attributed graphs plays an essential role in applications such as fraud detection, intrusion monitoring, and misinformation analysis. However, text-attributed graphs (TAGs), in which node information is expressed in natural language, remain underexplored, largely due to the absence of standardized benchmark datasets. In this work, we introduce TAG-AD, a comprehensive benchmark for anomaly node detection on TAGs. TAG-AD leverages large language models (LLMs) to generate realistic anomalous node texts directly in the raw text space, producing anomalies that are semantically coherent yet contextually inconsistent and thus more reflective of real-world irregularities. In addition, TAG-AD incorporates multiple other anomaly types, enabling thorough and reproducible evaluation of graph anomaly detection (GAD) methods. With these datasets, we further benchmark existing unsupervised GNN-based GAD methods as well as zero-shot LLMs for GAD. As part of our zero-shot detection setup, we propose a retrieval-augmented generation (RAG)-assisted, LLM-based zero-shot anomaly detection framework. The framework mitigates reliance on brittle, hand-crafted prompts by constructing a global anomaly knowledge base and distilling it into reusable analysis frameworks. Our experimental results reveal a clear division of strengths: LLMs are particularly effective at detecting contextual anomalies, whereas GNN-based methods remain superior for structural anomaly detection. Moreover, RAG-assisted prompting achieves performance comparable to human-designed prompts while eliminating manual prompt engineering, underscoring the practical value of our RAG-assisted zero-shot LLM anomaly detection framework.

cs.LG↗
arXiv2025

Stable and Efficient Charging of Superconducting Capacitively Shunted Flux Quantum Batteries

Quantum batteries, as miniature energy storage devices, have sparked significant research interest in recent years. However, achieving rapid and stable energy transfer in quantum batteries while obeying quantum speed limits remains a critical challenge. In this work, we experimentally optimize the charging process by leveraging the unique energy level structure of a superconducting capacitively-shunted flux qubit, using counterdiabatic pulses in the stimulated Raman adiabatic passage. Compared to previous studies, we impose two different norm constraints on the driving Hamiltonian, achieving optimal charging without exceeding the overall driving strength. Furthermore, we experimentally demonstrate a charging process that achieves the quantum speed limit. In addition, we introduce a dimensionless parameter $\mathcal{S}$ to unify charging speed and stability, offering a universal metric for performance optimization. In contrast to metrics such as charging power and thermodynamic efficiency, the $\mathcal{S}$ criterion quantitatively captures the stability of ergentropy while also considering the charging speed. Our results highlight the potential of the capacitively-shunted qubit platform as an ideal candidate for realizing three-level quantum batteries and deliver novel strategies for optimizing energy transfer protocols.

quant-ph↗
arXiv2025

Unconditional quantum teleportational advantage of single photons

Photon loss is the biggest enemy in quantum communications. Direct transmission typically results in a small fraction of photons surviving over long distances, hindering advanced applications such as loophole-free Bell tests and device-independent quantum key distribution. Quantum teleportation offers a "disembodied" way of transferring particles through a virtual channel--quantum entanglement--which, in principle, could overcome the photon transmission loss. Experimentally, however, no previous quantum teleportation of single photons has shown a higher survival probability than using direct transmission. To overcome this challenge, here we first propose and demonstrate an all-optical scheme for remote preparation of entangled photons. Through an effective 15-dB channel loss, we realize a heralding efficiency of 82% for event-ready entangled photons. Based on the prior distributed entanglement, we demonstrate that teleportation-based transmission efficiency offers a 2.95-fold enhancement, compared to using direct transmission through the same channel, therefore establishing an unconditional teleportational advantage.

quant-ph↗
arXiv2025

Many-body chaos and pole-skipping in holographic charged rotating fluids

Recent developments identify pole-skipping as a `smoking-gun' signature of the hydrodynamic nature of chaos, offering an alternative way to probe quantum chaos in addition to the out-of-time-ordered correlator (OTOC). We study the quantum chaos and pole-skipping phenomenon in the strongly coupled charged rotating fluids, holographically dual to rotating black holes with nontrivial gauge field. We find that the near-horizon equation governing energy-density fluctuations differs from the source-less shock wave equation determining the OTOC, which depends on the $U(1)$ gauge choice. This discrepancy is eliminated under an appropriate boundary condition on the $U(1)$ gauge potential at the event horizon, as required by the vanishing of Wilson loop at the Euclidean horizon. We further investigate the dependence of the butterfly velocity on the charge and rotation parameters in a specific black hole configuration--the Cvetič-Lü-Pope solution.

hep-th↗
arXiv2025

Improving DF-Conformer Using Hydra For High-Fidelity Generative Speech Enhancement on Discrete Codec Token

The Dilated FAVOR Conformer (DF-Conformer) is an efficient variant of the Conformer architecture designed for speech enhancement (SE). It employs fast attention through positive orthogonal random features (FAVOR+) to mitigate the quadratic complexity associated with self-attention, while utilizing dilated convolution to expand the receptive field. This combination results in impressive performance across various SE models. In this paper, we propose replacing FAVOR+ with bidirectional selective structured state-space sequence models to achieve two main objectives:(1) enhancing global sequential modeling by eliminating the approximations inherent in FAVOR+, and (2) maintaining linear complexity relative to the sequence length. Specifically, we utilize Hydra, a bidirectional extension of Mamba, framed within the structured matrix mixer framework. Experiments conducted using a generative SE model on discrete codec tokens, known as Genhancer, demonstrate that the proposed method surpasses the performance of the DF-Conformer.

cs.SD↗
arXiv2025

Experimental Quantum Channel Purification

Quantum networks, which integrate multiple quantum computers and the channels connecting them, are crucial for distributed quantum information processing but remain inherently susceptible to channel noise. Channel purification emerges as a promising technique for suppressing noise in quantum channels without complex encoding and decoding operations, making it particularly suitable for remote quantum information transmission in optical systems. In this work, we introduce an experimental setup for efficient channel purification, harnessing the spatial and polarization properties of photons. Our design employs two Fredkin gates to enable coherent interference between independent noise channels, achieving effective noise suppression across a wide range of noise levels and types. Through application to entanglement distribution, our protocol demonstrates a superior capability to preserve entanglement against channel noise compared to conventional entanglement purification methods.

quant-ph↗
arXiv2025

Entanglement Superactivation in Multiphoton Distillation Networks

In quantum networks, after passing through noisy channels or information processing, residual states may lack sufficient entanglement for further tasks, yet they may retain hidden quantum resources that can be recycled. Efficiently recycling these states to extract entanglement resources such as genuine multipartite entanglement or Einstein-Podolsky-Rosen pairs is essential for optimizing network performance. Here, we develop a tripartite entanglement distillation scheme using an eight-photon quantum platform, demonstrating entanglement superactivation phenomena which are unique to multipartite systems. We successfully generate a three-photon genuinely entangled state from two bi-separable states via local operations and classical communication, demonstrating superactivation of genuine multipartite entanglement. Furthermore, we extend our scheme to generate a three-photon state capable of extracting an Einstein-Podolsky-Rosen pair from two initial states lacking this capability, revealing a previously unobserved entanglement superactivation phenomenon. Our methods and findings offer not only practical applications for quantum networks, but also lead to a deeper understanding of multipartite entanglement structures.

quant-ph↗
arXiv2025

Multi-Agent Scenario Generation in Roundabouts with a Transformer-enhanced Conditional Variational Autoencoder

With the increasing integration of intelligent driving functions into serial-produced vehicles, ensuring their functionality and robustness poses greater challenges. Compared to traditional road testing, scenario-based virtual testing offers significant advantages in terms of time and cost efficiency, reproducibility, and exploration of edge cases. We propose a Transformer-enhanced Conditional Variational Autoencoder (CVAE-T) model for generating multi-agent traffic scenarios in roundabouts, which are characterized by high vehicle dynamics and complex layouts, yet remain relatively underexplored in current research. The results show that the proposed model can accurately reconstruct original scenarios and generate realistic, diverse synthetic scenarios. Besides, two Key-Performance-Indicators (KPIs) are employed to evaluate the interactive behavior in the generated scenarios. Analysis of the latent space reveals partial disentanglement, with several latent dimensions exhibiting distinct and interpretable effects on scenario attributes such as vehicle entry timing, exit timing, and velocity profiles. The results demonstrate the model's capability to generate scenarios for the validation of intelligent driving functions involving multi-agent interactions, as well as to augment data for their development and iterative improvement.

cs.RO↗