arXiv subjects

Zihao Li

Publications and source records attributed to Zihao Li.

At least 127 records · Page 7Linked to original sources

arXiv2025

AIM 2025 Challenge on High FPS Motion Deblurring: Methods and Results

This paper presents a comprehensive review of the AIM 2025 High FPS Non-Uniform Motion Deblurring Challenge, highlighting the proposed solutions and final results. The objective of this challenge is to identify effective networks capable of producing clearer and visually compelling images in diverse and challenging conditions, by learning representative visual cues for complex aggregations of motion types. A total of 68 participants registered for the competition, and 9 teams ultimately submitted valid entries. This paper thoroughly evaluates the state-of-the-art advances in high-FPS single image motion deblurring, showcasing the significant progress in the field, while leveraging samples of the novel dataset, MIORe, that introduces challenging examples of movement patterns.

cs.CV↗

arXiv2025

APEX$^2$: Adaptive and Extreme Summarization for Personalized Knowledge Graphs

Knowledge graphs (KGs), which store an extensive number of relational facts, serve various applications. Recently, personalized knowledge graphs (PKGs) have emerged as a solution to optimize storage costs by customizing their content to align with users' specific interests within particular domains. In the real world, on one hand, user queries and their underlying interests are inherently evolving, requiring PKGs to adapt continuously; on the other hand, the summarization is constantly expected to be as small as possible in terms of storage cost. However, the existing PKG summarization methods implicitly assume that the user's interests are constant and do not shift. Furthermore, when the size constraint of PKG is extremely small, the existing methods cannot distinguish which facts are more of immediate interest and guarantee the utility of the summarized PKG. To address these limitations, we propose APEX$^2$, a highly scalable PKG summarization framework designed with robust theoretical guarantees to excel in adaptive summarization tasks with extremely small size constraints. To be specific, after constructing an initial PKG, APEX$^2$ continuously tracks the interest shift and adjusts the previous summary. We evaluate APEX$^2$ under an evolving query setting on benchmark KGs containing up to 12 million triples, summarizing with compression ratios $\leq 0.1\%$. The experiments show that APEX outperforms state-of-the-art baselines in terms of both query-answering accuracy and efficiency. Code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/APEX.

cs.LG↗

arXiv2025

A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems

Recent advances in large language models have sparked growing interest in AI agents capable of solving complex, real-world tasks. However, most existing agent systems rely on manually crafted configurations that remain static after deployment, limiting their ability to adapt to dynamic and evolving environments. To this end, recent research has explored agent evolution techniques that aim to automatically enhance agent systems based on interaction data and environmental feedback. This emerging direction lays the foundation for self-evolving AI agents, which bridge the static capabilities of foundation models with the continuous adaptability required by lifelong agentic systems. In this survey, we provide a comprehensive review of existing techniques for self-evolving agentic systems. Specifically, we first introduce a unified conceptual framework that abstracts the feedback loop underlying the design of self-evolving agentic systems. The framework highlights four key components: System Inputs, Agent System, Environment, and Optimisers, serving as a foundation for understanding and comparing different strategies. Based on this framework, we systematically review a wide range of self-evolving techniques that target different components of the agent system. We also investigate domain-specific evolution strategies developed for specialised fields such as biomedicine, programming, and finance, where optimisation objectives are tightly coupled with domain constraints. In addition, we provide a dedicated discussion on the evaluation, safety, and ethical considerations for self-evolving agentic systems, which are critical to ensuring their effectiveness and reliability. This survey aims to provide researchers and practitioners with a systematic understanding of self-evolving AI agents, laying the foundation for the development of more adaptive, autonomous, and lifelong agentic systems.

cs.AI↗

arXiv2025

Randomized coordinate gradient descent almost surely escapes strict saddle points

We analyze the behavior of randomized coordinate gradient descent for nonconvex optimization, proving that under standard assumptions, the iterates almost surely escape strict saddle points. By formulating the method as a nonlinear random dynamical system and characterizing neighborhoods of critical points, we establish this result through the center-stable manifold theorem.

math.OC↗

arXiv2025

Enhancing Systematic Interoperability: Convergences and Mismatches between Web 3.0 and the EU Data Act

Interoperability is increasingly recognised as a foundational principle for fostering innovation, competition, and user autonomy in the evolving digital ecosystem. Existing research on interoperability predominantly focuses either on technological interoperability itself or on the legal regulations concerning interoperability, with insufficient exploration of their interdisciplinary intersection. This paper compares the technological interoperability in Web 3.0 with the theoretical framework of legal interoperability established by the EU Data Act, analysing the areas of convergence and mismatch. The goal is to align technical interoperability with legal concepts of interoperability, thereby enhancing the practical implementation of systematic interoperability in the next generation of the Web. This study finds that, firstly, Web 3.0's concept of interoperability spans data, systems, and applications, while the Data Act focuses solely on data. This narrow scope risks creating a fragmented ecosystem, where data exchange is possible, but full integration of systems and applications is hindered, leading to inefficiencies, and obstructing seamless data flow across platforms. Secondly, while Web 3.0 technically seeks to achieve interoperability through the integration of entire systems and decentralised applications, the compliance with Data Act might negatively limit such system and application interoperability through its data interoperability provisions. This paper suggests interdisciplinary recommendations to enhance the implementation and enforcement of interoperability. On one hand, the Data Act should broaden its concept of interoperability to encompass both the systems and applications layers. On the other hand, it is advisable to introduce provisions for standardised protocols through soft law mechanisms to address legal shortcomings and keep pace with technological advancements.

cs.CY↗

arXiv2025

Efficient certification of high-dimensional entanglement

High-dimensional entanglement (HDE) is a valuable resource in quantum information processing, and efficient certification of HDE is crucial to many applications. In this work, we propose a simple and general framework for certifying HDE in general bipartite pure states under restricted operations, such as local operations and classical communication (LOCC). On this basis we show that HDE in general bipartite pure states can be certified efficiently. Moreover, the sample cost for certifying a given degree of HDE even decreases monotonically with the local dimensions. In addition, for a general two-qubit pure state, we construct an optimal entanglement certification strategy based on separable operations, which can be realized by LOCC when the target state has sufficiently high entanglement. The core concept of our framework is versatile and can be extended to certify a wide range of critical resources under restricted operations.

quant-ph↗

arXiv2025

A 13-Billion-Year View of Galaxy Growth: Metallicity Gradient Evolution from the Local Universe to $z=9$ with JWST and Archival Surveys

The galaxy gas-phase metallicity gradients have been extensively studied over the past four decades, both in the local and high-redshift universe, as they trace the baryon cycle and growth of galaxies. With the unprecedented spatial resolution and sensitivity of JWST, it is now possible to measure metallicity and its radial gradients out to redshifts as high as $z = 9$. Here, we present a sample of 455 spectroscopically confirmed galaxies from redshifts $1.7 \lesssim z \lesssim 9$ that are spatially resolved on sub-kiloparsec (kpc) scales by deep JWST NIRCam or NIRISS Wide Field Slitless Spectroscopy (WFSS). Synthesizing these new JWST observations with legacy observations from the literature, we observe that at redshift $z > 5$, galaxy centers are more metal-rich, exhibiting negative metallicity gradients of $\sim-0.4$ dex kpc$^{-1}$. These gradients flatten over time, reaching near-zero around $z \approx 2$, coinciding with the peak of the cosmic star formation rate. Beyond this point, the gradients become negative again at lower redshifts approaching $z=0$. This evolution likely reflects transitions in galaxy formation modes: an inside-out growth phase dominated by intense central star formation with inefficient feedback and limited gas mixing during ``cosmic dawn", enhanced gas mixing due to feedback-driven wind and gas accretion at ``cosmic noon", and a later phase of slow evolution and reduced feedback toward the present day. These physical processes, including gas accretion and feedback, not only regulate star and galaxy formation on a cosmic scale but also shape the evolutionary pathways of individual galaxies over cosmic time.

astro-ph.GA↗

arXiv2025

Logarithmic Comparison-Based Query Complexity for Fair Division of Indivisible Goods

We study the problem of fairly allocating $m$ indivisible goods to $n$ agents, where agents may have different preferences over the goods. In the traditional setting, agents' valuations are provided as inputs to the algorithm. In this paper, we adopt the query model, which has been widely considered for other similar problems (such as matching [Nis21], graph isomorphism [OS18], and equilibrium in game [Bab16]), and apply it to the fair division problem. In particular, we consider a new \emph{comparison-based query model}, where the algorithm presents two bundles of goods to an agent and the agent responds by telling the algorithm which bundle she prefers. We investigate the query complexity for computing allocations with several fairness notions, including \emph{proportionality up to one good} (PROP1), \emph{envy-freeness up to one good} (EF1), and \emph{maximin share} (MMS). Our main result is an algorithm that computes an allocation that satisfies both PROP1 and $\frac12$-MMS within $O(\log m)$ queries with a constant number of $n$ agents. For identical and additive valuation, we present an algorithm for computing an EF1 allocation within $O(\log m)$ queries with a constant number of $n$ agents. To complement the positive results, we show that the lower bound of the query complexity for any of the three fairness notions is $Ω(\log m)$ even with two agents.

cs.GT↗

arXiv2025

U.S. Port Disruptions under Tropical Cyclones: Resilience Analysis by Harnessing Multiple-Source Dataset

This study introduces the CyPort Dataset, recording disruptions to 145 U.S. principal ports and freight network from 90 tropical cyclones (2015-2023). It addresses limitations of event specific resilience studies and provides a comprehensive dataset for broader analysis. To account for excess zeros and unobserved heterogeneity in disruption outcomes, the Random Parameter Negative Binomial Lindley (RPNB Lindley) model is employed to produce more reliable resilience insights. The model demonstrates improved fit over traditional methods and uncovers variation in how features such as wind speed, storm surge height, rainfall, and distance to cyclone influence disruption outcomes across ports. This analysis reveals a tipping point at Saffir Simpson Hurricane Category 4, where disruptions escalate sharply, causing greater impacts and prolonged recovery. Regionally, ports along the Gulf of America show greatest vulnerability. Within the freight network, ports with high betweenness centrality are more resilient, while transshipment and local hubs are more fragile.

stat.AP↗

arXiv2025

Graph and Sequential Neural Networks in Session-based Recommendation: A Survey

Recent years have witnessed the remarkable success of recommendation systems (RSs) in alleviating the information overload problem. As a new paradigm of RSs, session-based recommendation (SR) specializes in users' short-term preference capture and aims to provide a more dynamic and timely recommendation based on the ongoing interacted actions. In this survey, we will give a comprehensive overview of the recent works on SR. First, we clarify the definitions of various SR tasks and introduce the characteristics of session-based recommendation against other recommendation tasks. Then, we summarize the existing methods in two categories: sequential neural network based methods and graph neural network (GNN) based methods. The standard frameworks and technical are also introduced. Finally, we discuss the challenges of SR and new research directions in this area.

cs.IR↗

arXiv2025

Feature Extraction and Steering for Enhanced Chain-of-Thought Reasoning in Language Models

Large Language Models (LLMs) demonstrate the ability to solve reasoning and mathematical problems using the Chain-of-Thought (CoT) technique. Expanding CoT length, as seen in models such as DeepSeek-R1, significantly enhances this reasoning for complex problems, but requires costly and high-quality long CoT data and fine-tuning. This work, inspired by the deep thinking paradigm of DeepSeek-R1, utilizes a steering technique to enhance the reasoning ability of an LLM without external datasets. Our method first employs Sparse Autoencoders (SAEs) to extract interpretable features from vanilla CoT. These features are then used to steer the LLM's internal states during generation. Recognizing that many LLMs do not have corresponding pre-trained SAEs, we further introduce a novel SAE-free steering algorithm, which directly computes steering directions from the residual activations of an LLM, obviating the need for an explicit SAE. Experimental results demonstrate that both our SAE-based and subsequent SAE-free steering algorithms significantly enhance the reasoning capabilities of LLMs.

cs.CL↗

arXiv2025

Listwise Preference Alignment Optimization for Tail Item Recommendation

Preference alignment has achieved greater success on Large Language Models (LLMs) and drawn broad interest in recommendation research. Existing preference alignment methods for recommendation either require explicit reward modeling or only support pairwise preference comparison. The former directly increases substantial computational costs, while the latter hinders training efficiency on negative samples. Moreover, no existing effort has explored preference alignment solutions for tail-item recommendation. To bridge the above gaps, we propose LPO4Rec, which extends the Bradley-Terry model from pairwise comparison to listwise comparison, to improve the efficiency of model training. Specifically, we derive a closed form optimal policy to enable more efficient and effective training without explicit reward modeling. We also present an adaptive negative sampling and reweighting strategy to prioritize tail items during optimization and enhance performance in tail-item recommendations. Besides, we theoretically prove that optimizing the listwise preference optimization (LPO) loss is equivalent to maximizing the upper bound of the optimal reward. Our experiments on three public datasets show that our method outperforms 10 baselines by a large margin, achieving up to 50% performance improvement while reducing 17.9% GPU memory usage when compared with direct preference optimization (DPO) in tail-item recommendation. Our code is available at https://github.com/Yuhanleeee/LPO4Rec.

cs.IR↗

arXiv2025

Developing ChemDFM as a large language foundation model for chemistry

Artificial intelligence (AI) has played an increasingly important role in chemical research. However, most models currently used in chemistry are specialist models that require training and tuning for specific tasks. A more generic and efficient solution would be an AI model that could address many tasks and support free-form dialogue in the broad field of chemistry. In its utmost form, such a generalist AI chemist could be referred to as Chemical General Intelligence. Large language models (LLMs) have recently logged tremendous success in the general domain of natural language processing, showing emerging task generalization and free-form dialogue capabilities. However, domain knowledge of chemistry is largely missing when training general-domain LLMs. The lack of such knowledge greatly hinders the performance of generalist LLMs in the field of chemistry. To this end, we develop ChemDFM, a pioneering LLM for chemistry trained on 34B tokens from chemical literature and textbooks, and fine-tuned using 2.7M instructions. As a result, it can understand and reason with chemical knowledge in free-form dialogue. Quantitative evaluations show that ChemDFM significantly surpasses most representative open-source LLMs. It outperforms GPT-4 on a great portion of chemical tasks, despite the substantial size difference. We have open-sourced the inference codes, evaluation datasets, and model weights of ChemDFM on Huggingface (https://huggingface.co/OpenDFM/ChemDFM-v1.0-13B).

cs.CL↗

arXiv2025

FORGE: An LLM-driven Framework for Large-Scale Smart Contract Vulnerability Dataset Construction

High-quality smart contract vulnerability datasets are critical for evaluating security tools and advancing smart contract security research. Two major limitations of current manual dataset construction are (1) labor-intensive and error-prone annotation processes limiting the scale, quality, and evolution of the dataset, and (2) absence of standardized classification rules results in inconsistent vulnerability categories and labeling results across different datasets. To address these limitations, we present FORGE, the first automated approach for constructing smart contract vulnerability datasets. FORGE leverages an LLM-driven pipeline to extract high-quality vulnerabilities from real-world audit reports and classify them according to the CWE, the most widely recognized classification in software security. FORGE employs a divide-and-conquer strategy to extract structured and self-contained vulnerability information from these reports. Additionally, it uses a tree-of-thoughts technique to classify the vulnerability information into the hierarchical CWE classification. To evaluate FORGE's effectiveness, we run FORGE on 6,454 real-world audit reports and generate a dataset comprising 81,390 solidity files and 27,497 vulnerability findings across 296 CWE categories. Manual assessment of the dataset demonstrates high extraction precision and classification consistency with human experts (precision of 95.6% and inter-rater agreement k-$α$ of 0.87). We further validate the practicality of our dataset by benchmarking 13 existing security tools on our dataset. The results reveal the significant limitations in current detection capabilities. Furthermore, by analyzing the severity-frequency distribution patterns through a unified CWE perspective in our dataset, we highlight inconsistency between current smart contract research focus and priorities identified from real-world vulnerabilities...

cs.CR↗

arXiv2025

Multi-Interest Recommendation: A Survey

Existing recommendation methods often struggle to model users' multifaceted preferences due to the diversity and volatility of user behavior, as well as the inherent uncertainty and ambiguity of item attributes in practical scenarios. Multi-interest recommendation addresses this challenge by extracting multiple interest representations from users' historical interactions, enabling fine-grained preference modeling and more accurate recommendations. It has drawn broad interest in recommendation research. However, current recommendation surveys have either specialized in frontier recommendation methods or delved into specific tasks and downstream applications. In this work, we systematically review the progress, solutions, challenges, and future directions of multi-interest recommendation by answering the following three questions: (1) Why is multi-interest modeling significantly important for recommendation? (2) What aspects are focused on by multi-interest modeling in recommendation? and (3) How can multi-interest modeling be applied, along with the technical details of the representative modules? We hope that this survey establishes a fundamental framework and delivers a preliminary overview for researchers interested in this field and committed to further exploration. The implementation of multi-interest recommendation summarized in this survey is maintained at https://github.com/WHUIR/Multi-Interest-Recommendation-A-Survey.

cs.IR↗

arXiv2025

Experimental Verification of Entangled States in the Adversarial Scenario

Efficient verification of entangled states is crucial to many applications in quantum information processing. However, the effectiveness of standard quantum state verification (QSV) is based on the condition of independent and identical distribution (IID), which impedes its applications in many practical scenarios. Here we demonstrate a defensive QSV protocol, which is effective in all kinds of non-IID scenarios, including the extremely challenging adversarial scenario. To this end, we build a high-speed preparation-and-measurement apparatus controlled by quantum random-number generators. Our experiments clearly show that standard QSV protocols often provide unreliable fidelity certificates in non-IID scenarios. In sharp contrast, the defensive QSV protocol based on a homogeneous strategy can provide reliable and nearly tight fidelity certificates at comparable high efficiency, even under malicious attacks. Moreover, our scheme is robust against the imperfections in a realistic experiment, which is very appealing to practical applications.

quant-ph↗

arXiv2025

Learnable Spatial-Temporal Positional Encoding for Link Prediction

Accurate predictions rely on the expressiveness power of graph deep learning frameworks like graph neural networks and graph transformers, where a positional encoding mechanism has become much more indispensable in recent state-of-the-art works to record the canonical position information. However, the current positional encoding is limited in three aspects: (1) most positional encoding methods use pre-defined, and fixed functions, which are inadequate to adapt to the complex attributed graphs; (2) a few pioneering works proposed the learnable positional encoding but are still limited to the structural information, not considering the real-world time-evolving topological and feature information; (3) most positional encoding methods are equipped with transformers' attention mechanism to fully leverage their capabilities, where the dense or relational attention is often unaffordable on large-scale structured data. Hence, we aim to develop Learnable Spatial-Temporal Positional Encoding in an effective and efficient manner and propose a simple temporal link prediction model named L-STEP. Briefly, for L-STEP, we (1) prove the proposed positional learning scheme can preserve the graph property from the spatial-temporal spectral viewpoint, (2) verify that MLPs can fully exploit the expressiveness and reach transformers' performance on that encoding, (3) change different initial positional encoding inputs to show robustness, (4) analyze the theoretical complexity and obtain less empirical running time than SOTA, and (5) demonstrate its temporal link prediction out-performance on 13 classic datasets and with 10 algorithms in both transductive and inductive settings using 3 different sampling strategies. Also, L-STEP obtains the leading performance in the newest large-scale TGB benchmark. Our code is available at https://github.com/kthrn22/L-STEP.

cs.LG↗

arXiv2025

Fair Division with Allocator's Preference

We study the fair allocation of indivisible resources among agents. Most prior work focuses on fairness and/or efficiency among agents. However, the allocator, as the resource owner, may also be involved in many scenarios (e.g., government resource allocation, heritage division, company personnel assignment, etc). The allocator inclines to obtain a fair or efficient allocation based on her preference over the items and to whom each item is allocated. We propose a model and study two problems: 1) Find an allocation fair to both agents and allocator; 2) Maximize allocator's efficiency under agents' fairness. We consider the two fundamental fairness criteria: envy-freeness and proportionality. For the first problem, we study the existence of an allocation that is envy-free up to $c$ goods (EF-$c$) or proportional up to $c$ goods (PROP-$c$) from both the agents' and the allocator's perspectives, called doubly EF-$c$ or doubly PROP-$c$. When the allocator's utility only depends on the items (not recipients), we prove that a doubly EF-$1$ allocation always exists. For the general setting where the allocator has a preference over the items and to whom each item is allocated, a doubly EF-$1$ allocation always exists for two agents, a doubly PROP-$2$ allocation always exists for personalized bi-valued valuations, and a doubly PROP-$O(\log n)$ allocation always exists. For the second problem, we give (in)approximability results with asymptotically tight bounds in most settings. When agents' valuations are binary, maximizing the allocator's social welfare while ensuring agents' fairness criteria of PROP-$c$ (with a general number of agents) and EF-$c$ (with a constant number of agents) are both polynomial-time solvable for any integer $c$. Strong inapproximability holds for most of the other settings (general valuations, EF-$c$, etc).

cs.GT↗