arXiv subjects

Jun Zhao

Publications and source records attributed to Jun Zhao.

At least 109 records · Page 6Linked to original sources

arXiv2025

The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Models

Reasoning models have demonstrated exceptional performance in tasks such as mathematics and logical reasoning, primarily due to their ability to engage in step-by-step thinking during the reasoning process. However, this often leads to overthinking, resulting in unnecessary computational overhead. To address this issue, Mode Selection aims to automatically decide between Long-CoT (Chain-of-Thought) or Short-CoT by utilizing either a Thinking or NoThinking mode. Simultaneously, Early Exit determines the optimal stopping point during the iterative reasoning process. Both methods seek to reduce the computational burden. In this paper, we first identify Mode Selection as a more challenging variant of the Early Exit problem, as they share similar objectives but differ in decision timing. While Early Exit focuses on determining the best stopping point for concise reasoning at inference time, Mode Selection must make this decision at the beginning of the reasoning process, relying on pre-defined fake thoughts without engaging in an explicit reasoning process, referred to as zero-step thinking. Through empirical studies on nine baselines, we observe that prompt-based approaches often fail due to their limited classification capabilities when provided with minimal hand-crafted information. In contrast, approaches that leverage internal information generally perform better across most scenarios but still exhibit issues with stability. Our findings indicate that existing methods relying solely on the information provided by models are insufficient for effectively addressing Mode Selection in scenarios with limited information, highlighting the ongoing challenges of this task. Our code is available at https://github.com/Trae1ounG/Zero_Step_Thinking.

cs.AI↗

arXiv2025

Towards Agentic Self-Learning LLMs in Search Environment

We study whether self-learning can scale LLM-based agents without relying on human-curated datasets or predefined rule-based rewards. Through controlled experiments in a search-agent setting, we identify two key determinants of scalable agent training: the source of reward signals and the scale of agent task data. We find that rewards from a Generative Reward Model (GRM) outperform rigid rule-based signals for open-domain learning, and that co-evolving the GRM with the policy further boosts performance. Increasing the volume of agent task data-even when synthetically generated-substantially enhances agentic capabilities. Building on these insights, we propose \textbf{Agentic Self-Learning} (ASL), a fully closed-loop, multi-role reinforcement learning framework that unifies task generation, policy execution, and evaluation within a shared tool environment and LLM backbone. ASL coordinates a Prompt Generator, a Policy Model, and a Generative Reward Model to form a virtuous cycle of harder task setting, sharper verification, and stronger solving. Empirically, ASL delivers steady, round-over-round gains, surpasses strong RLVR baselines (e.g., Search-R1) that plateau or degrade, and continues improving under zero-labeled-data conditions, indicating superior sample efficiency and robustness. We further show that GRM verification capacity is the main bottleneck: if frozen, it induces reward hacking and stalls progress; continual GRM training on the evolving data distribution mitigates this, and a small late-stage injection of real verification data raises the performance ceiling. This work establishes reward source and data scale as critical levers for open-domain agent learning and demonstrates the efficacy of multi-role co-evolution for scalable, self-improving agents. The data and code of this paper are released at https://github.com/forangel2014/Towards-Agentic-Self-Learning

cs.AI↗

arXiv2025

PrivTuner with Homomorphic Encryption and LoRA: A P3EFT Scheme for Privacy-Preserving Parameter-Efficient Fine-Tuning of AI Foundation Models

AI foundation models have recently demonstrated impressive capabilities across a wide range of tasks. Fine-tuning (FT) is a method of customizing a pre-trained AI foundation model by further training it on a smaller, targeted dataset. In this paper, we initiate the study of the Privacy-Preserving Parameter-Efficient FT (P3EFT) framework, which can be viewed as the intersection of Parameter-Efficient FT (PEFT) and Privacy-Preserving FT (PPFT). PEFT modifies only a small subset of the model's parameters to achieve FT (i.e., adapting a pre-trained model to a specific dataset), while PPFT uses privacy-preserving technologies to protect the confidentiality of the model during the FT process. There have been many studies on PEFT or PPFT but very few on their fusion, which motivates our work on P3EFT to achieve both parameter efficiency and model privacy. To exemplify our P3EFT, we present the PrivTuner scheme, which incorporates Fully Homomorphic Encryption (FHE) enabled privacy protection into LoRA (short for ``Low-Rank Adapter''). Intuitively speaking, PrivTuner allows the model owner and the external data owners to collaboratively implement PEFT with encrypted data. After describing PrivTuner in detail, we further investigate its energy consumption and privacy protection. Then, we consider a PrivTuner system over wireless communications and formulate a joint optimization problem to adaptively minimize energy while maximizing privacy protection, with the optimization variables including FDMA bandwidth allocation, wireless transmission power, computational resource allocation, and privacy protection. A resource allocation algorithm is devised to solve the problem. Experiments demonstrate that our algorithm can significantly reduce energy consumption while adapting to different privacy requirements.

cs.CR↗

arXiv2025

Query2Triple: Unified Query Encoding for Answering Diverse Complex Queries over Knowledge Graphs

Complex Query Answering (CQA) is a challenge task of Knowledge Graph (KG). Due to the incompleteness of KGs, query embedding (QE) methods have been proposed to encode queries and entities into the same embedding space, and treat logical operators as neural set operators to obtain answers. However, these methods train KG embeddings and neural set operators concurrently on both simple (one-hop) and complex (multi-hop and logical) queries, which causes performance degradation on simple queries and low training efficiency. In this paper, we propose Query to Triple (Q2T), a novel approach that decouples the training for simple and complex queries. Q2T divides the training into two stages: (1) Pre-training a neural link predictor on simple queries to predict tail entities based on the head entity and relation. (2) Training a query encoder on complex queries to encode diverse complex queries into a unified triple form that can be efficiently solved by the pretrained neural link predictor. Our proposed Q2T is not only efficient to train, but also modular, thus easily adaptable to various neural link predictors that have been studied well. Extensive experiments demonstrate that, even without explicit modeling for neural set operators, Q2T still achieves state-of-the-art performance on diverse complex queries over three public benchmarks.

cs.AI↗

arXiv2025

MotivGraph-SoIQ: Integrating Motivational Knowledge Graphs and Socratic Dialogue for Enhanced LLM Ideation

Large Language Models (LLMs) hold substantial potential for accelerating academic ideation but face critical challenges in grounding ideas and mitigating confirmation bias for further refinement. We propose integrating motivational knowledge graphs and socratic dialogue to address these limitations in enhanced LLM ideation (MotivGraph-SoIQ). This novel framework provides essential grounding and practical idea improvement steps for LLM ideation by integrating a Motivational Knowledge Graph (MotivGraph) with a Q-Driven Socratic Ideator. The MotivGraph structurally stores three key node types(problem, challenge and solution) to offer motivation grounding for the LLM ideation process. The Ideator is a dual-agent system utilizing Socratic questioning, which facilitates a rigorous refinement process that mitigates confirmation bias and improves idea quality across novelty, experimental rigor, and motivational rationality dimensions. On the ICLR25 paper topics dataset, MotivGraph-SoIQ exhibits clear advantages over existing state-of-the-art approaches across LLM-based scoring, ELO ranking, and human evaluation metrics.

cs.CL↗

arXiv2025

Citrus-V: Advancing Medical Foundation Models with Unified Medical Image Grounding for Clinical Reasoning

Medical imaging provides critical evidence for clinical diagnosis, treatment planning, and surgical decisions, yet most existing imaging models are narrowly focused and require multiple specialized networks, limiting their generalization. Although large-scale language and multimodal models exhibit strong reasoning and multi-task capabilities, real-world clinical applications demand precise visual grounding, multimodal integration, and chain-of-thought reasoning. We introduce Citrus-V, a multimodal medical foundation model that combines image analysis with textual reasoning. The model integrates detection, segmentation, and multimodal chain-of-thought reasoning, enabling pixel-level lesion localization, structured report generation, and physician-like diagnostic inference in a single framework. We propose a novel multimodal training approach and release a curated open-source data suite covering reasoning, detection, segmentation, and document understanding tasks. Evaluations demonstrate that Citrus-V outperforms existing open-source medical models and expert-level imaging systems across multiple benchmarks, delivering a unified pipeline from visual grounding to clinical reasoning and supporting precise lesion quantification, automated reporting, and reliable second opinions.

cs.CV↗

arXiv2025

Direct Observation of d-Wave Superconducting Gap Symmetry in Pressurized La3Ni2O7-delta Single Crystals

The recent discovery of superconductivity in pressure-stabilized bulk La3Ni2O7-delta, with a critical temperature (Tc) exceeding 77 K, has opened a new frontier in high-temperature superconductivity research beyond cuprates. Yet, the superconducting gap amplitude and symmetry, the key parameters to characterize a superconductor, remain elusive due to the overwhelming challenges of gap studies under high pressure. Here, we introduce in situ directional point-contact spectroscopy conducted under truly hydrostatic pressure, enabling the direct mapping of the superconducting gap in pressurized La3Ni2O7-delta single crystals. Depending on the junction orientation, differential conductance (dI/dV) spectra exhibit distinct V-shaped quasiparticle features and a sharp zero-bias peak, indicating a predominant d-wave-like pairing symmetry. Measurement of the c-axis gap amplitude Delta yields a gap-to-Tc ratio of 2Delta/kBTc = 4.2(5), positioning La3Ni2O7-delta firmly among unconventional, nodal high-Tc superconductors. These findings set stringent constraints on theoretical models for nickelate superconductors and establish a robust spectroscopic approach for understanding superconductors under extreme pressures.

cond-mat.supr-con↗

arXiv2025

Joint Optimization of Energy Consumption and Completion Time in Federated Learning

Federated Learning (FL) is an intriguing distributed machine learning approach due to its privacy-preserving characteristics. To balance the trade-off between energy and execution latency, and thus accommodate different demands and application scenarios, we formulate an optimization problem to minimize a weighted sum of total energy consumption and completion time through two weight parameters. The optimization variables include bandwidth, transmission power and CPU frequency of each device in the FL system, where all devices are linked to a base station and train a global model collaboratively. Through decomposing the non-convex optimization problem into two subproblems, we devise a resource allocation algorithm to determine the bandwidth allocation, transmission power, and CPU frequency for each participating device. We further present the convergence analysis and computational complexity of the proposed algorithm. Numerical results show that our proposed algorithm not only has better performance at different weight parameters (i.e., different demands) but also outperforms the state of the art.

cs.LG↗

arXiv2025

PL-CA: A Parametric Legal Case Augmentation Framework

Conventional RAG is considered one of the most effective methods for addressing model knowledge insufficiency and hallucination, particularly in the judicial domain that requires high levels of knowledge rigor, logical consistency, and content integrity. However, the conventional RAG method only injects retrieved documents directly into the model's context, which severely constrains models due to their limited context windows and introduces additional computational overhead through excessively long contexts, thereby disrupting models' attention and degrading performance on downstream tasks. Moreover, many existing benchmarks lack expert annotation and focus solely on individual downstream tasks while real-world legal scenarios consist of multiple mixed legal tasks, indicating conventional benchmarks' inadequacy for reflecting models' true capabilities. To address these limitations, we propose PL-CA, which introduces a parametric RAG (P-RAG) framework to perform data augmentation on corpus knowledge and encode this legal knowledge into parametric vectors, and then integrates this parametric knowledge into the LLM's feed-forward networks (FFN) via LoRA, thereby alleviating models' context pressure. Additionally, we also construct a multi-task legal dataset comprising more than 2000 training and test instances, which are all expert-annotated and manually verified. We conduct our experiments on our dataset, and the experimental results demonstrate that our method reduces the overhead associated with excessively long contexts while maintaining competitive performance on downstream tasks compared to conventional RAG. Our code and dataset are provided in the appendix.

cs.CL↗

arXiv2025

An LLM-enabled semantic-centric framework to consume privacy policies

In modern times, people have numerous online accounts, but they rarely read the Terms of Service or Privacy Policy of those sites, despite claiming otherwise, due to the practical difficulty in comprehending them. The mist of data privacy practices forms a major barrier for user-centred Web approaches, and for data sharing and reusing in an agentic world. Existing research proposed methods for using formal languages and reasoning for verifying the compliance of a specified policy, as a potential cure for ignoring privacy policies. However, a critical gap remains in the creation or acquisition of such formal policies at scale. We present a semantic-centric approach for using state-of-the-art large language models (LLM), to automatically identify key information about privacy practices from privacy policies, and construct $\mathit{Pr}^2\mathit{Graph}$, knowledge graph with grounding from Data Privacy Vocabulary (DPV) for privacy practices, to support downstream tasks. Along with the pipeline, the $\mathit{Pr}^2\mathit{Graph}$ for the top-100 popular websites is also released as a public resource, by using the pipeline for analysis. We also demonstrate how the $\mathit{Pr}^2\mathit{Graph}$ can be used to support downstream tasks by constructing formal policy representations such as Open Digital Right Language (ODRL) or perennial semantic Data Terms of Use (psDToU). To evaluate the technology capability, we enriched the Policy-IE dataset by employing legal experts to create custom annotations. We benchmarked the performance of different large language models for our pipeline and verified their capabilities. Overall, they shed light on the possibility of large-scale analysis of online services' privacy practices, as a promising direction to audit the Web and the Internet. We release all datasets and source code as public resources to facilitate reuse and improvement.

cs.AI↗

arXiv2025

Semantic and Temporal Integration in Latent Diffusion Space for High-Fidelity Video Super-Resolution

Recent advancements in video super-resolution (VSR) models have demonstrated impressive results in enhancing low-resolution videos. However, due to limitations in adequately controlling the generation process, achieving high fidelity alignment with the low-resolution input while maintaining temporal consistency across frames remains a significant challenge. In this work, we propose Semantic and Temporal Guided Video Super-Resolution (SeTe-VSR), a novel approach that incorporates both semantic and temporal-spatio guidance in the latent diffusion space to address these challenges. By incorporating high-level semantic information and integrating spatial and temporal information, our approach achieves a seamless balance between recovering intricate details and ensuring temporal coherence. Our method not only preserves high-reality visual content but also significantly enhances fidelity. Extensive experiments demonstrate that SeTe-VSR outperforms existing methods in terms of detail recovery and perceptual quality, highlighting its effectiveness for complex video super-resolution tasks.

cs.CV↗

arXiv2025

Let's Measure the Elephant in the Room: Facilitating Personalized Automated Analysis of Privacy Policies at Scale

In modern times, people have numerous online accounts, but they rarely read the Terms of Service or Privacy Policy of those sites despite claiming otherwise. This paper introduces PoliAnalyzer, a neuro-symbolic system that assists users with personalized privacy policy analysis. PoliAnalyzer uses Natural Language Processing (NLP) to extract formal representations of data usage practices from policy texts. In favor of deterministic, logical inference is applied to compare user preferences with the formal privacy policy representation and produce a compliance report. To achieve this, we extend an existing formal Data Terms of Use policy language to model privacy policies as app policies and user preferences as data policies. In our evaluation using our enriched PolicyIE dataset curated by legal experts, PoliAnalyzer demonstrated high accuracy in identifying relevant data usage practices, achieving F1-score of 90-100% across most tasks. Additionally, we demonstrate how PoliAnalyzer can model diverse user data-sharing preferences, derived from prior research as 23 user profiles, and perform compliance analysis against the top 100 most-visited websites. This analysis revealed that, on average, 95.2% of a privacy policy's segments do not conflict with the analyzed user preferences, enabling users to concentrate on understanding the 4.8% (636 / 13205) that violates preferences, significantly reducing cognitive burden. Further, we identified common practices in privacy policies that violate user expectations - such as the sharing of location data with 3rd parties. This paper demonstrates that PoliAnalyzer can support automated personalized privacy policy analysis at scale using off-the-shelf NLP tools. This sheds light on a pathway to help individuals regain control over their data and encourage societal discussions on platform data practices to promote a fairer power dynamic.

cs.CL↗

arXiv2025

QuHE: Optimizing Utility-Cost in Quantum Key Distribution and Homomorphic Encryption Enabled Secure Edge Computing Networks

Ensuring secure and efficient data processing in mobile edge computing (MEC) systems is a critical challenge. While quantum key distribution (QKD) offers unconditionally secure key exchange and homomorphic encryption (HE) enables privacy-preserving data processing, existing research fails to address the comprehensive trade-offs among QKD utility, HE security, and system costs. This paper proposes a novel framework integrating QKD, transciphering, and HE for secure and efficient MEC. QKD distributes symmetric keys, transciphering bridges symmetric encryption, and HE processes encrypted data at the server. We formulate an optimization problem balancing QKD utility, HE security, processing and wireless transmission costs. However, the formulated optimization is non-convex and NPhard. To solve it efficiently, we propose the Quantum-enhanced Homomorphic Encryption resource allocation (QuHE) algorithm. Theoretical analysis proves the proposed QuHE algorithm's convergence and optimality, and simulations demonstrate its effectiveness across multiple performance metrics.

cs.SI↗

arXiv2025

Enhancing Data Processing Efficiency in Blockchain Enabled Metaverse over Wireless Communications

In the rapidly evolving landscape of the Metaverse, enhanced by blockchain technology, the efficient processing of data has emerged as a critical challenge, especially in wireless communication systems. Addressing this challenge, our paper introduces the innovative concept of data processing efficiency (DPE), aiming to maximize processed bits per unit of resource consumption in blockchain-empowered Metaverse environments. To achieve this, we propose the DPE-Aware User Association and Resource Allocation (DAUR) algorithm, a tailored optimization framework for blockchain-enabled Metaverse wireless communication systems characterized by joint computing and communication resource constraints. The DAUR algorithm transforms the nonconvex problem of maximizing the sum of DPE ratios into a solvable convex optimization problem. It alternates the optimization of key variables, including user association, work offloading ratios, task-specific computing resource distribution, bandwidth allocation, user power usage ratios, and server computing resource allocation ratios. Our extensive numerical results demonstrate the DAUR algorithm's effectiveness in DPE.

eess.SP↗

arXiv2025

JoyTTS: LLM-based Spoken Chatbot With Voice Cloning

JoyTTS is an end-to-end spoken chatbot that combines large language models (LLM) with text-to-speech (TTS) technology, featuring voice cloning capabilities. This project is built upon the open-source MiniCPM-o and CosyVoice2 models and trained on 2000 hours of conversational data. We have also provided the complete training code to facilitate further development and optimization by the community. On the testing machine seed-tts-zh, it achieves a SS (speaker similarity) score of 0.73 and a WER (Word Error Rate) of 5.09. The code and models, along with training and inference scripts, are available at https://github.com/jdh-algo/JoyTTS.git.

cs.SD↗

arXiv2025

InSight-R: A Framework for Risk-informed Human Failure Event Identification and Interface-Induced Risk Assessment Driven by AutoGraph

Human reliability remains a critical concern in safety-critical domains such as nuclear power, where operational failures are often linked to human error. While conventional human reliability analysis (HRA) methods have been widely adopted, they rely heavily on expert judgment for identifying human failure events (HFEs) and assigning performance influencing factors (PIFs). This reliance introduces challenges related to reproducibility, subjectivity, and limited integration of interface-level data. In particular, current approaches lack the capacity to rigorously assess how human-machine interface design contributes to operator performance variability and error susceptibility. To address these limitations, this study proposes a framework for risk-informed human failure event identification and interface-induced risk assessment driven by AutoGraph (InSight-R). By linking empirical behavioral data to the interface-embedded knowledge graph (IE-KG) constructed by the automated graph-based execution framework (AutoGraph), the InSight-R framework enables automated HFE identification based on both error-prone and time-deviated operational paths. Furthermore, we discuss the relationship between designer-user conflicts and human error. The results demonstrate that InSight-R not only enhances the objectivity and interpretability of HFE identification but also provides a scalable pathway toward dynamic, real-time human reliability assessment in digitalized control environments. This framework offers actionable insights for interface design optimization and contributes to the advancement of mechanism-driven HRA methodologies.

cs.HC↗

arXiv2025

A Troublemaker with Contagious Jailbreak Makes Chaos in Honest Towns

With the development of large language models, they are widely used as agents in various fields. A key component of agents is memory, which stores vital information but is susceptible to jailbreak attacks. Existing research mainly focuses on single-agent attacks and shared memory attacks. However, real-world scenarios often involve independent memory. In this paper, we propose the Troublemaker Makes Chaos in Honest Town (TMCHT) task, a large-scale, multi-agent, multi-topology text-based attack evaluation framework. TMCHT involves one attacker agent attempting to mislead an entire society of agents. We identify two major challenges in multi-agent attacks: (1) Non-complete graph structure, (2) Large-scale systems. We attribute these challenges to a phenomenon we term toxicity disappearing. To address these issues, we propose an Adversarial Replication Contagious Jailbreak (ARCJ) method, which optimizes the retrieval suffix to make poisoned samples more easily retrieved and optimizes the replication suffix to make poisoned samples have contagious ability. We demonstrate the superiority of our approach in TMCHT, with 23.51%, 18.95%, and 52.93% improvements in line topology, star topology, and 100-agent settings. Encourage community attention to the security of multi-agent systems.

cs.CL↗

arXiv2025

Agent-RewardBench: Towards a Unified Benchmark for Reward Modeling across Perception, Planning, and Safety in Real-World Multimodal Agents

As Multimodal Large Language Models (MLLMs) advance, multimodal agents show promise in real-world tasks like web navigation and embodied intelligence. However, due to limitations in a lack of external feedback, these agents struggle with self-correction and generalization. A promising approach is to use reward models as external feedback, but there is no clear on how to select reward models for agents. Thus, there is an urgent need to build a reward bench targeted at agents. To address these challenges, we propose Agent-RewardBench, a benchmark designed to evaluate reward modeling ability in MLLMs. The benchmark is characterized by three key features: (1) Multiple dimensions and real-world agent scenarios evaluation. It covers perception, planning, and safety with 7 scenarios; (2) Step-level reward evaluation. It allows for the assessment of agent capabilities at the individual steps of a task, providing a more granular view of performance during the planning process; and (3) Appropriately difficulty and high-quality. We carefully sample from 10 diverse models, difficulty control to maintain task challenges, and manual verification to ensure the integrity of the data. Experiments demonstrate that even state-of-the-art multimodal models show limited performance, highlighting the need for specialized training in agent reward modeling. Code is available at github.

cs.CL↗