arXiv subjects

Zhen Tao

Publications and source records attributed to Zhen Tao.

At least 19 recordsLinked to original sources

arXiv2026

SoK: From Generation to Consumption of Privacy Documents in Software Systems

Privacy documents (e.g., privacy policies) are a central mechanism through which digital services disclose data practices and seek user consent. Over the past decades, research on privacy documents has expanded significantly, encompassing not only traditional privacy policies but also short notices (e.g., privacy labels) and interface-level transparency mechanisms. As this research area continues to grow, it has become increasingly difficult to obtain a coherent view of how privacy documents are created, analyzed, evaluated, and maintained across their lifecycle. This SoK provides a unified, lifecycle-oriented view of privacy documents from a software engineering perspective. We systematically review and analyze 290 papers published between 2010 and 2025, organizing them around five research questions that examine how privacy documents are (1) defined and scoped, (2) generated, (3) analyzed and extracted, (4) checked for inconsistencies and noncompliance, and (5) evaluated and improved for usability. Building on our findings, we identify 15 key research trends and 21 open opportunities. We further chart four broader research directions that highlight (i) emerging challenges in AI-centric platforms, (ii) the need for diverse and up-to-date data foundations, (iii) LLM-based unified policy-code analysis, and (iv) dual usability for end-users and developers. We hope this SoK provides a shared foundation for future research on privacy policies and privacy documents.

cs.CR↗

arXiv2026

A Perspective on Phase Space Electronic Structure Theory : From Its Surface Hopping Origins Through To Its Future Promise

We trace the history of phase space electronic structure theory (PSEST), high- lighting how this powerful approach emerged from fundamental questions in semi- classical surface hopping dynamics and evolved into an alternative to standard Born- Oppenheimer based electronic structure theory (with moving instead of frozen nuclei). Our goal herein is not to recapitulate the mathematical details of phase space electronic structure calculations, and few equations are presented so as to maximize readability. Instead, our goal is to provide intuition for those new to the field both (i) regarding the physics present when solving the Schrodinger equation in a non-inertial frame as well as (ii) regarding why PSEST is a necessary step forward towards understanding chemical problems involving spin (with limited practical alternatives). We further high- light some of the many open questions in this fast developing area, which will hopefully inspire new practitioners in this field. This intuitive perspective lacks many equations and is meant to complement (rather than replace) the more technical review given in Bian et al, Chem. Phys. Rev. 7, 011303 (2026)

physics.chem-ph↗

arXiv2026

Fishing Out Free Riders: Shapley-Based Reward Attribution for Parallel Reasoning via Reinforcement Learning

Large Language Models (LLMs) excel at multi-step reasoning, yet current parallel reasoning approaches often fail to distinguish the contributions of individual reasoning paths. Many paths may be redundant, misleading, or even detrimental, but outcome-level rewards assign uniform reward, leading to ambiguous learning signals and unstable training. We propose Parallel Shapley, a reinforcement learning framework that attributes fine-grained, path-level contributions in multi-path reasoning. Treating each path as a player in a cooperative game, we leverage Shapley values to quantify marginal contributions, using a generative reward model to evaluate path utilities and Monte Carlo sampling for efficient approximation. Experiments on mathematical reasoning benchmarks show that Parallel Shapley outperforms existing baselines while providing more stable and interpretable training. Our framework effectively "fishes out the free riders," assigning reward proportionally and improving multi-path reasoning in LLMs.

cs.AI↗

arXiv2026

MemConflict: Evaluating Long-Term Memory Systems Under Memory Conflicts

Long-term memory systems enable conversational agents based on large language models (LLMs) to retain, retrieve, and apply user-specific information across multi-session interactions. However, existing evaluations mainly assess outcome-level performance or temporal updating, providing limited insight into how systems retrieve and rank temporally valid, factually correct, and contextually applicable memory evidence under conflicting alternatives. To address this gap, we propose MemConflict, a diagnostic framework that treats memory validity as a query-conditioned fitness-for-use problem. MemConflict formalizes dynamic, static, and conditional conflicts over temporal validity, factual correctness, and contextual applicability. It simulates controlled long-horizon histories from structured user profiles, introduces cross-session conflicts, and injects semantically similar distractors to create competition among memory candidates. The resulting multi-session dialogue benchmark supports black-box evaluation of final answers and white-box analysis of supporting-memory retrieval and ranking. Experiments on six representative long-term memory systems show uneven strengths across conflict types, with answer correctness often diverging from memory retrieval and ranking. Sensitivity analyses reveal that longer histories, distractors, implicit queries, and larger conflict distances degrade performance. Diagnostics show failures from missing supporting memories and ineffective use of retrieved memories. Collectively, MemConflict advances principled long-term memory governance through retrieval-aware, conflict-aware reliability assessment.

cs.IR↗

arXiv2026

SAGER: Self-Evolving User Policy Skills for Recommendation Agent

Large language model (LLM) based recommendation agents personalize what they know through evolving per-user semantic memory, yet how they reason remains a universal, static system prompt shared identically across all users. This asymmetry is a fundamental bottleneck: when a recommendation fails, the agent updates its memory of user preferences but never interrogates the decision logic that produced the failure, leaving its reasoning process structurally unchanged regardless of how many mistakes it accumulates. To address this bottleneck, we propose SAGER (Self-Evolving Agent for Personalized Recommendation), the first recommendation agent framework in which each user is equipped with a dedicated policy skill, a structured natural-language document encoding personalized decision principles that evolves continuously through interaction. SAGER introduces a two-representation skill architecture that decouples a rich evolution substrate from a minimal inference-time injection, an incremental contrastive chain-of-thought engine that diagnoses reasoning flaws by contrasting accepted against unchosen items while preserving accumulated priors, and skill-augmented listwise reasoning that creates fine-grained decision boundaries where the evolved skill provides genuine discriminative value. Experiments on four public benchmarks demonstrate that SAGER achieves state-of-the-art performance, with gains orthogonal to memory accumulation, confirming that personalizing the reasoning process itself is a qualitatively distinct source of recommendation improvement.

cs.IR↗

arXiv2026

MemEmo: Evaluating Emotion in Memory Systems of Agents

Memory systems address the challenge of context loss in Large Language Model during prolonged interactions. However, compared to human cognition, the efficacy of these systems in processing emotion-related information remains inconclusive. To address this gap, we propose an emotion-enhanced memory evaluation benchmark to assess the performance of mainstream and state-of-the-art memory systems in handling affective information. We developed the \textbf{H}uman-\textbf{L}ike \textbf{M}emory \textbf{E}motion (\textbf{HLME}) dataset, which evaluates memory systems across three dimensions: emotional information extraction, emotional memory updating, and emotional memory question answering. Experimental results indicate that none of the evaluated systems achieve robust performance across all three tasks. Our findings provide an objective perspective on the current deficiencies of memory systems in processing emotional memories and suggest a new trajectory for future research and system optimization.

cs.CL↗

arXiv2026

Ab Initio Random Matrix Theory of Molecular Electronic Structure

We use ab initio electronic-structure methods to investigate random-matrix theory (RMT) universality in molecular electronic structure. Using single-reference electronic structure methods, including Hartree-Fock, configuration-interaction singles (CIS), density functional theory, and linear-response time-dependent density-functional theory, we compute single-particle orbital energies and many-electron excitations of several representative molecules (benzene, alanine, 1-phenylethylamine, methyloxirane, and helicene chains). For generic low-symmetry geometries, the unfolded spectra of these ab initio Hamiltonians exhibit Wigner-Dyson level statistics of the Gaussian orthogonal ensemble (GOE). For extended helicene chains we explicitly restrict to bound valence excitations below the ionization threshold and still observe GOE statistics, indicating that the RMT universality is present for physical states of direct relevance to real molecules. We further explore the electric and magnetic field dependence of the molecular electronic spectra. The variance of electric polarizability (level curvature K) is predicted to be non-analytic in the magnetic field which serves as an infrared cutoff, proportional to log(1/|B|). We observe a transition to the Gaussian unitary ensemble (GUE) by increasing the magnetic fields, although it occurs only at magnetic fields far beyond experimentally accessible scales. Our results indicate that random matrix universality provides a general framework for organizing ab initio predictions of interacting electron spectra in complex systems.

cond-mat.str-el↗

arXiv2026

Charge Transfer with a Spin. I: A Generalized CASSCF Framework for Investigating Charge Transfer in the Presence of Spin-Orbit Coupling

We present a generalized extension of the recently developed electron/hole-transfer Dynamically-weighted State-Averaged Constrained CASSCF (eDSC/hDSC) method to model charge transfer in the presence of spin-orbit coupling (SOC) for systems containing an odd number of electrons. Our approach incorporates complex-valued spinor orbitals and incorporates four electronic configurations in describing ground-excited state curve crossings between Kramers-restricted doublet states. The method achieves smooth potential energy surfaces and rapid self-consistent field (SCF) convergence across a wide range of spin-orbit coupling strengths, providing an efficient framework for investigating charge transfer processes in the presence of nontrivial spin degrees of freedom.

physics.chem-ph↗

arXiv2026

Charge Transfer with a Spin. II: A Framework for Diabatization which Localizes Charge and Spin

We investigate a diabatization procedure that localizes charges (in real space) and localizes spins (in spin space) for open-shell systems that exhibit charge transfer in the presence of spin-orbit coupling. The procedure is applied to a two-state crossing between pairs of Kramers-restricted doublet states (which can also be considered effectively a four-state crossing). To generate the relevant electronic states, we employ the recently developed electron/hole-transfer Dynamically-weighted State-Averaged Constrained CASSCF (eDSC/hDSC) method that treats systems with an odd number of electrons. To generate the relevant diabatic states, we employ a two-step optimization over complex-unitary rotations that sequentially maximizes dipole and spin moments through iterative Jacobi sweeps; the resulting update rules are effectively equivalent to those of approximate joint diagonalization (AJD) applied to charge and spin. The method converges rapidly and yields smooth diabatic potential energy surfaces that preserve dipole and spin properties (e.g., a slowly varying pseudospin texture) along the reaction coordinate while maintaining time-reversal symmetry.

physics.chem-ph↗

arXiv2025

A Longitudinal Measurement of Privacy Policy Evolution for Large Language Models

Large language model (LLM) services have been rapidly integrated into people's daily lives as chatbots and agentic systems. They are nourished by collecting rich streams of data, raising privacy concerns around excessive collection of sensitive personal information. Privacy policies are the fundamental mechanism for informing users about data practices in modern information privacy paradigm. Although traditional web and mobile policies are well studied, the privacy policies of LLM providers, their LLM-specific content, and their evolution over time remain largely underexplored. In this paper, we present the first longitudinal empirical study of privacy policies for mainstream LLM providers worldwide. We curate a chronological dataset of 74 historical privacy policies and 115 supplemental privacy documents from 11 LLM providers across 5 countries up to August 2025, and extract over 3,000 sentence-level edits between consecutive policy versions. We compare LLM privacy policies to those of other software formats, propose a taxonomy tailored to LLM privacy policies, annotate policy edits and align them with a timeline of key LLM ecosystem events. Results show they are substantially longer, demand college-level reading ability, and remain highly vague. Our taxonomy analysis reveals patterns in how providers disclose LLM-specific practices and highlights regional disparities in coverage. Policy edits are concentrated in first-party data collection and international/specific-audience sections, and that product releases and regulatory actions are the primary drivers, shedding light on the status quo and the evolution of LLM privacy policies.

cs.CR↗

arXiv2025

Task-Aware Retrieval Augmentation for Dynamic Recommendation

Dynamic recommendation systems aim to provide personalized suggestions by modeling temporal user-item interactions across time-series behavioral data. Recent studies have leveraged pre-trained dynamic graph neural networks (GNNs) to learn user-item representations over temporal snapshot graphs. However, fine-tuning GNNs on these graphs often results in generalization issues due to temporal discrepancies between pre-training and fine-tuning stages, limiting the model's ability to capture evolving user preferences. To address this, we propose TarDGR, a task-aware retrieval-augmented framework designed to enhance generalization capability by incorporating task-aware model and retrieval-augmentation. Specifically, TarDGR introduces a Task-Aware Evaluation Mechanism to identify semantically relevant historical subgraphs, enabling the construction of task-specific datasets without manual labeling. It also presents a Graph Transformer-based Task-Aware Model that integrates semantic and structural encodings to assess subgraph relevance. During inference, TarDGR retrieves and fuses task-aware subgraphs with the query subgraph, enriching its representation and mitigating temporal generalization issues. Experiments on multiple large-scale dynamic graph datasets demonstrate that TarDGR consistently outperforms state-of-the-art methods, with extensive empirical evidence underscoring its superior accuracy and generalization capabilities.

cs.IR↗

arXiv2025

Dynamic Graph Recommendation via Sparse Augmentation and Singular Adaptation

Dynamic recommendation, focusing on modeling user preference from historical interactions and providing recommendations on current time, plays a key role in many personalized services. Recent works show that pre-trained dynamic graph neural networks (GNNs) can achieve excellent performance. However, existing methods by fine-tuning node representations at large scales demand significant computational resources. Additionally, the long-tail distribution of degrees leads to insufficient representations for nodes with sparse interactions, posing challenges for efficient fine-tuning. To address these issues, we introduce GraphSASA, a novel method for efficient fine-tuning in dynamic recommendation systems. GraphSASA employs test-time augmentation by leveraging the similarity of node representation distributions during hierarchical graph aggregation, which enhances node representations. Then it applies singular value decomposition, freezing the original vector matrix while focusing fine-tuning on the derived singular value matrices, which reduces the parameter burden of fine-tuning and improves the fine-tuning adaptability. Experimental results demonstrate that our method achieves state-of-the-art performance on three large-scale datasets.

cs.SI↗

arXiv2025

Non-Resonant Raman Optical Activity From Phase-Space Electronic Structure Theory

In order to model experimental non-resonant Raman optical activity, chemists must compute a host of second-order response tensors, (e.g. the electric-dipole magnetic-dipole polarizability) and their nuclear derivatives along a set of vibrational modes. While these response functions are almost always computed within a Born-Oppenheimer (BO) framework, here we provide a natural interpretation of the electric-dipole magnetic-dipole polarizability within phase space electronic structure theory, a beyond-BO model whereby the electronic structure depends on nuclear momentum (P) in addition to nuclear position (R). By coupling to nuclear momentum, phase space electronic structure theory is able to capture the asymmetric response of the electronic properties to an external field, in sofar as for a vibrating (non-stationary) molecule, dmu/dB \ne dm/dF, where mu and m are the electrical linear and magnetic dipoles, and F and B are electric and magnetic fields. As an example, for a prototypical methyloxirane molecule, we show that phase space electronic structure theory is able to deliver a reasonably good match with experimental results in a manner that is invariant to gauge origin G0.

physics.chem-ph↗

arXiv2025

dInfer: An Efficient Inference Framework for Diffusion Language Models

Diffusion-based large language models (dLLMs) have emerged as a promising alternative to autoregressive (AR) LLMs, leveraging denoising-based generation to enable inherent parallelism. Even more and more open-sourced dLLM models emerge, yet their widespread adoption remains constrained by the lack of a standardized and efficient inference framework. We present dInfer, an efficient and extensible framework for dLLM inference. dInfer decomposes the inference pipeline into four modular components--model, diffusion iteration manager, decoding strategy, and KV-cache manager--and integrates novel algorithms for each component alongside system-level optimizations. Through this combination of algorithmic innovations and system enhancements, dInfer achieves substantial efficiency gains without compromising output quality on LLaDA-MoE. At batch size 1, it surpasses 1,100 tokens per second on HumanEval and averages over 800 tokens per second across six benchmarks on $8\times$ H800 GPUs. Compared to prior systems, dInfer delivers a $10\times$ speedup over Fast-dLLM while maintaining similar model performance. Even compared to the AR model (with a comparable number of activation parameters and performance) QWen2.5-3B, which is highly optimized with the latest vLLM inference engine, dInfer still delivers a $2$-$3\times$ speedup. The implementation of dInfer is open-sourced at https://github.com/inclusionAI/dInfer.

cs.CL↗

arXiv2025

Towards Context-aware Mobile Privacy Notice: Implementation of A Deployable Contextual Privacy Policies Generator

Lengthy and legally phrased privacy policies impede users' understanding of how mobile applications collect and process personal data. Prior work proposed Contextual Privacy Policies (CPPs) for mobile apps to display shorter policy snippets only in the corresponding user interface contexts, but the pipeline could not be deployable in real-world mobile environments. In this paper, we present PrivScan, the first deployable CPP Software Development Kit (SDK) for Android. It captures live app screenshots to identify GUI elements associated with types of personal data and displays CPPs in a concise, user-facing format. We provide a lightweight floating button that offers low-friction, on-demand control. The architecture leverages remote deployment to decouple the multimodal backend pipeline from a mobile client comprising five modular components, thereby reducing on-device resource demands and easing cross-platform portability. A feasibility-oriented evaluation shows an average execution time of 9.15\,s, demonstrating the practicality of our approach. The source code of PrivScan is available at https://github.com/buyanghc/PrivScan and the demo video can be found at https://www.youtube.com/watch?v=ck-25otfyHc.

cs.CR↗

arXiv2025

MemOS: A Memory OS for AI System

Large Language Models (LLMs) have become an essential infrastructure for Artificial General Intelligence (AGI), yet their lack of well-defined memory management systems hinders the development of long-context reasoning, continual personalization, and knowledge consistency.Existing models mainly rely on static parameters and short-lived contextual states, limiting their ability to track user preferences or update knowledge over extended periods.While Retrieval-Augmented Generation (RAG) introduces external knowledge in plain text, it remains a stateless workaround without lifecycle control or integration with persistent representations.Recent work has modeled the training and inference cost of LLMs from a memory hierarchy perspective, showing that introducing an explicit memory layer between parameter memory and external retrieval can substantially reduce these costs by externalizing specific knowledge. Beyond computational efficiency, LLMs face broader challenges arising from how information is distributed over time and context, requiring systems capable of managing heterogeneous knowledge spanning different temporal scales and sources. To address this challenge, we propose MemOS, a memory operating system that treats memory as a manageable system resource. It unifies the representation, scheduling, and evolution of plaintext, activation-based, and parameter-level memories, enabling cost-efficient storage and retrieval. As the basic unit, a MemCube encapsulates both memory content and metadata such as provenance and versioning. MemCubes can be composed, migrated, and fused over time, enabling flexible transitions between memory types and bridging retrieval with parameter-based learning. MemOS establishes a memory-centric system framework that brings controllability, plasticity, and evolvability to LLMs, laying the foundation for continual learning and personalized modeling.

cs.CL↗

arXiv2025

The Phase-Space Way To Electronic Structure Theory and Subsequently Chemical Dynamics

Phase-space electronic structure theory offers up a new and powerful approach for tackling problems with coupled nuclear-electronic dynamics in a fashion that goes beyond Born-Oppenheimer (BO) theory. Whereas BO theory stipulates that we consider electronic states parameterized by nuclear position $X$ only, i.e. molecular orbitals are functions of nuclear positions but not nuclear velocities, phase-space (PS) theory allows for electronic states to be parameterized by both nuclear position X and nuclear momentum $P$. As a result, within a phase-space approach, one can directly recover many new features, including electronic momentum and vibrational circular dichroism spectra. Moreover, phase-space electronic structure theory is exact for the hydrogen atom and, for a set of model problems, the method can even improve upon vibrational energies relative to BO theory. Perhaps most importantly, the phase-space approach offers up a very new perspective on spin physics, stipulating that molecules and materials with degenerate or nearly degenerate ground states (due to spin degeneracy) display broken-symmetry ground states in their phase-space potential energy surfaces. This last feature opens up very new possibilities for exploring spin chemistry (including the Einstein-de Haas effect and chiral induced spin selectivity) within the context of more established electronic structure theory. At the end of the day, in order to tackle electronic dynamical phenomena, especially subtle problems in magnetic chemistry, it will be essential for the electronic structure community to pivot towards diagonalizing $\hat H_{PS}(X, P)$ rather than $\hat H_{BO}(X)$.

physics.chem-ph↗

arXiv2025

Conical Intersections and Electronic Momentum As Viewed From Phase Space Electronic Structure Theory

We investigate the structure of a prototypical two-state conical intersection (BeH$_2$) using a phase space electronic Hamiltonian $\hat{H}_{PS}(\bR,\bP)$ that goes beyond the Born-Oppenheimer framework. By parameterizing the electronic Schr{\"o}dinger equation by both nuclear position ($\bR$) and momentum ($\bP$), we solve for quantum electronic states in a moving frame that can break time reversal symmetry and, as a result, the branching plane of the conical intersection within a phase space framework now has dimension three (rather than dimension two as found within the standard Born-Oppenheimer framework). Moreover, we note that, if one fixes a geometry in real space that lies in the conical intersection seam and scans over the corresponding momentum space, one finds a double well (with minima at $\pm \bP_{min} \ne 0$), indicating that the stationary electronic states of the phase space electronic Hamiltonian carry electronic momentum -- a feature that cannot be captured by a Born-Oppenheimer electronic state. Interestingly, for $BeH_2$, this electronic momenta (as calculated with full configuration interaction) agrees with what is predicted by approximate complex restricted Hartree-Fock calculations, indicating a physical interpretation of complex instabilities in modern electronic structure calculations. Altogether, this study suggests that we have still have a lot to learn about conical intersections when it comes to electronic momentum, and highlights the urgent need for more experiments to probe what photochemical observables can and/or cannot be captured by standard electronic structure that isolates conical intersections within the Born-Oppenheimer framework.

physics.chem-ph↗