Searcharxiv⌕ Search

arXiv subjects

Nan Zhang

Publications and source records attributed to Nan Zhang.

At least 37 records · Page 2Linked to original sources
arXiv2026

Cascade of magnetic-field-induced quantum spin states in a spin-1 honeycomb magnet

Quantum fluctuations endow spin systems with surprisingly enriched magnetic phase diagrams. In frustrated magnets, strong quantum fluctuations boosted by either geometrical incompatibility or competitive exchange interactions stabilize cascades of unusual phases of matter. Here we reveal the presence of multiple quantum phases in the honeycomb antiferromagnet Na$_{3}$Ni$_{2}$BiO$_{6}$, both inside and beyond its field-induced one-third magnetization plateau. Comprehensive measurements of thermodynamic quantities demonstrate that the one-third plateau comprises at least three distinct spin states with nearly-degenerate net magnetization, separated by first-order transitions that likely involve sequential spin reconfiguration. Upon further increasing the magnetic field, the system evolves across a myriad of peculiar phases before reaching full polarization; these intermediate phases possess copious low-energy excitations, manifested by anomalous upturns of specific heat at ultralow temperatures -- probably hinting at the development of "hidden" ordered ground states. The complex magnetic phase diagram of Na$_{3}$Ni$_{2}$BiO$_{6}$ underlines the preponderant impact of quantum fluctuations on a honeycomb spin lattice with strong exchange frustration.

cond-mat.str-el↗
arXiv2026

TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning

This paper investigates large language model (LLM) abstention learning, specifically using ternary reward, which incentivize truthfulness in large language models. This paper extends that idea by moving from a ternary reward to a Trajectory-Informed advantage reweighting, dynamically re-weights the abstention reward during Group Relative Policy Optimization (GRPO) training. The objective of this work focuses on abstention learning instead of improving truthfulness, serving as an exploration into hallucination reduction. The novelty of this paper lies in methodological innovation, advantage re-weighting, and benchmark selection. Leveraging GRPO's multiple trajectories as a natural abstention signal, this method uses a reward signal to explore knowledge boundaries and encourage consistency. By demonstrating that trajectories can be used as a confidence indicator of the policy relative to the query, they are then used to dynamically calculate the abstention advantage. AbstentionBench is used as the evaluation benchmark, as this work aims to contribute to the field of abstention learning. All datasets on the benchmark were tested against this method and various baselines. Empirical results demonstrate that TIAR achieves state-of-the-art abstention F1 scores across five of six evaluation categories, outperforming the static ternary baseline on 17 of 31 benchmark datasets while fully preserving baseline accuracy.

cs.CL↗
arXiv2026

Mental Damage: Caption Poisoning Attacks on Retrieval-Augmented Text-to-Music Generation

Retrieval-augmented text-to-music (TTM) systems augment underspecified user prompts using captions retrieved from a music caption dataset. This design introduces an integrity dependency on the music knowledge database. We show that an attacker can poison the database by injecting a small number of crafted music captions, causing the system to retrieve malicious captions that bias prompt augmentation and steer generation away from the user's intended function, without modifying the user prompt, retriever, or generator. To achieve the music caption poisoning attack, we propose a dual-layer caption poisoning strategy that preserves high-level retrieval anchors while injecting low-level acoustic descriptors to steer prompt augmentation and downstream music generation toward an attacker-chosen target intent. In a MusicCaps knowledge database, CLAP retriever, and MusicGen pipeline, poisoned generations move substantially closer to the attacker's target, while remaining comparably aligned with the original user query. These results expose a practical integrity risk for retrieval-augmented creative AI systems. Our demo can be found at: https://yizhu-wen.github.io/Mental-Damage/

cs.SD↗
arXiv2026

Position: Generative Engine Optimization Creates Underexamined Risks, Governance Must Target Concentration, Disclosure, and Academic Blind Spots

Large language model (LLM) answer engines are increasingly used for information seeking, shifting visibility from ranked lists to synthesized answers. This enables Generative Engine Optimization (GEO), which targets LLM answer engines' evidence pool and generation. We analyze the search engine optimization (SEO) to GEO transition to identify two risks: (i) concentrated influence from low contestability and system sensitivity, and (ii) undisclosed commercial influence embedded in evidence and reasoning. We then formalize a general GEO pipeline to locate where optimization acts and compare academic and industry practices, revealing a third risk: (iii) academic-industry blind spots driven by visibility and evaluation asymmetries between offline setups and deployed systems. This position argues the need for answer-level governance and measurement: stronger contestability, high-precision disclosure, black-box auditing of material influence, and deployment-aligned metrics for exposure persistence.

cs.CY↗
arXiv2026

An Asymptotic-Preserving Dual Formulation Finite-Volume Method for the Thermal Rotating Shallow Water Equations

We propose a new second-order asymptotic-preserving (AP) dual formulation finite-volume (DF-FV) method for the thermal rotating shallow water (TRSW) equations. The TRSW system models geophysical flows characterized by horizontal temperature/density variations, exhibiting multi-scale dynamics due to the coexistence of fast rotational waves and slower advective processes. To efficiently address challenges associated with the multiscale nature of the TRSW system, we follow the DF-FV framework and develop a DF-FV method, in which both the conservative and nonconservative (primitive) forms of the equations are simultaneously solved, allowing the method to exploit the complementary strengths of each representation across different flow regimes. The primitive formulation is better suited for preserving the correct asymptotic behavior in nearly thermal quasi-geostrophic (TQG) regimes characterized by a low Rossby number, while the conservative formulation is essential for robust shock capturing in high-Rossby-number regimes, in which nonconservative discretizations may fail to converge to physically relevant weak solutions.

math.NA↗
arXiv2026

Efficient PRM Training Data Synthesis via Formal Verification

Process Reward Models (PRMs) have emerged as a promising approach for improving LLM reasoning capabilities by providing process supervision over reasoning traces. However, existing approaches for constructing PRM training data remain costly and noisy, as they typically rely on human annotation or sampling-based labeling methods that require repeated LLM calls. In this work, we propose FoVer, a framework that synthesizes PRM training data from formal reasoning tasks by annotating step-level error labels using formal verification tools such as Z3 and Isabelle. By leveraging formal verification, FoVer enables efficient and accurate PRM data construction without requiring human annotation or additional LLM calls. Using FoVer, we create PRM training data from formal logic and theorem proving tasks. Experiments on 12 reasoning benchmarks show that fine-tuning on our training data improves PRMs not only on math and logic reasoning tasks, which are informal variants of the training tasks, but also on NLI and BBH benchmarks, which differ substantially from the tasks used to construct the training data. These results demonstrate the practical effectiveness of FoVer, showing that PRM training data created using formal verification improves PRMs on informal reasoning tasks written in natural language. The datasets, models, and code are provided at https://github.com/psunlpgroup/FoVer.

cs.CL↗
arXiv2026

TwinLoop: Simulation-in-the-Loop Digital Twins for Online Multi-Agent Reinforcement Learning

Decentralised online learning enables runtime adaptation in cyber-physical multi-agent systems, but when operating conditions change, learned policies often require substantial trial-and-error interaction before recovering performance. To address this, we propose TwinLoop, a simulation-in-the-loop digital twin framework for online multi-agent reinforcement learning. When a context shift occurs, the digital twin is triggered to reconstruct the current system state, initialise from the latest agent policies, and perform accelerated policy improvement with simulation what-if analysis before synchronising updated parameters back to the agents in the physical system. We evaluate TwinLoop in a vehicular edge computing task-offloading scenario with changing workload and infrastructure conditions. The results suggest that digital twins can improve post-shift adaptation efficiency and reduce reliance on costly online trial-and-error.

cs.LG↗
arXiv2026

BAAI Cardiac Agent: An intelligent multimodal agent for automated reasoning and diagnosis of cardiovascular diseases from cardiac magnetic resonance imaging

Cardiac magnetic resonance (CMR) is a cornerstone for diagnosing cardiovascular disease. However, it remains underutilized due to complex, time-consuming interpretation across multi-sequences, phases, quantitative measures that heavily reliant on specialized expertise. Here, we present BAAI Cardiac Agent, a multimodal intelligent system designed for end-to-end CMR interpretation. The agent integrates specialized cardiac expert models to perform automated segmentation of cardiac structures, functional quantification, tissue characterization and disease diagnosis, and generates structured clinical reports within a unified workflow. Evaluated on CMR datasets from two hospitals (2413 patients) spanning 7-types of major cardiovascular diseases, the agent achieved an area under the receiver-operating-characteristic curve exceeding 0.93 internally and 0.81 externally. In the task of estimating left ventricular function indices, the results generated by this system for core parameters such as ejection fraction, stroke volume, and left ventricular mass are highly consistent with clinical reports, with Pearson correlation coefficients all exceeding 0.90. The agent outperformed state-of-the-art models in segmentation and diagnostic tasks, and generated clinical reports showing high concordance with expert radiologists (six readers across three experience levels). By dynamically orchestrating expert models for coordinated multimodal analysis, this agent framework enables accurate, efficient CMR interpretation and highlights its potentials for complex clinical imaging workflows. Code is available at https://github.com/plantain-herb/Cardiac-Agent.

eess.IV↗
arXiv2026

GMPilot: An Expert AI Agent For FDA cGMP Compliance

The pharmaceutical industry is facing challenges with quality management such as high costs of compliance, slow responses and disjointed knowledge. This paper presents GMPilot, a domain-specific AI agent that is designed to support FDA cGMP compliance. GMPilot is based on a curated knowledge base of regulations and historical inspection observations and uses Retrieval-Augmented Generation (RAG) and Reasoning-Acting (ReAct) frameworks to provide real-time and traceable decision support to the quality professionals. In a simulated inspection scenario, GMPilot shows how it can improve the responsiveness and professionalism of quality professionals by providing structured knowledge retrieval and verifiable regulatory and case-based support. Although GMPilot lacks in the aspect of regulatory scope and model interpretability, it is a viable avenue of improving quality management decision-making in the pharmaceutical sector using intelligent approaches and an example of specialized application of AI in highly regulated sectors.

cs.AI↗
arXiv2026

Maintenance optimization of a two-component system with mixed observability

This paper studies maintenance optimization for a two-component system under mixed observability. Component~$U_1$ is fully monitored, whereas component~$U_2$ is only partially observable due to sensing limitations. The system exhibits unidirectional positive degradation dependence, in which the health state of component~$U_1$ influences the degradation process of component~$U_2$, but not vice versa. We propose a novel framework for modeling and optimizing maintenance decisions for such systems using a partially observable Markov decision process (POMDP). Under mild conditions, we analytically establish structural properties of the optimal maintenance policy. Baum-Welch algorithm with multiple sample paths is developed to estimate the unknown system parameters in the context of a covariate-dependent Hidden Markov Model. %from observational data with multiple trajectories. Numerical experiments demonstrate the effectiveness of the proposed parameter estimation and the maintenance policy. Across 64 instances, we show that it consistently outperforms classical threshold-based policies. Specifically, when the degradation of component $U_1$ is faster, it achieves maximal cost reductions of up to approximately $6\%$

math.OC↗
arXiv2026

When Reasoning Meets Compression: Understanding the Effects of LLMs Compression on Large Reasoning Models

Compression methods, including quantization, distillation, and pruning, improve the computational efficiency of large reasoning models (LRMs). However, existing studies either fail to sufficiently compare all three compression methods on LRMs or lack in-depth interpretation analysis. In this paper, we investigate how the reasoning capabilities of LRMs are compromised during compression, through performance benchmarking and mechanistic interpretation. To uncover the effects of compression on reasoning performance, we benchmark quantized, distilled, and pruned DeepSeek-R1 models on four reasoning datasets (AIME 2024, FOLIO, Temporal Sequences, and MuSiQue). To precisely locate compression effects on model weights, we adapt difference of means and attribution patching techniques, focusing on the activation of every linear component in compressed LRMs, to interpret fine-grained causal relationships between weights and various reasoning capabilities. This fine-grained interpretation addresses a fundamental question of compression: which weights are the most important for reasoning? Overall, we find dynamically quantized 2.51-bit R1 reaches close-to-R1 performance. With empirical verification, we present three main findings that generalize across both R1 and non-R1 LRMs: (1) Weight count has a greater impact on LRMs' knowledge memorization than reasoning, highlighting the risks of pruning and distillation; (2) The MLP up projection in the final layer of distilled LRMs is one of the most important components, offering a new perspective on locating critical weights - a fundamental problem in model compression; and (3) Current quantization methods overly compress the final-layer modules and MLP gate projections, so protecting just 2% of all weights that are excessively compressed can raise average accuracy by 6.57%, greatly surpassing the state-of-the-art.

cs.LG↗
arXiv2026

S$^3$GND: An Effective Learning-Based Approach for Subgraph Similarity Search Under Generalized Neighbor Difference Semantics (Technical Report)

Subgraph similarity search over large-scale graphs is a fundamental task that retrieves subgraphs similar to a given query graph from a data graph, and it plays a crucial role in real applications such as protein discovery, social network analysis, and recommendation systems. While prior works on subgraph similarity search studied various graph similarity metrics, in this paper, we propose a novel graph similarity semantics, \textit{generalized neighbor difference} (GND), that accounts for both the keyword-set relationships between vertices and edge-weight differences. We formulate the problem of \textit{subgraph similarity search under the generalized neighbor difference semantics} (S$^3$GND), which retrieves those subgraphs similar to a query graph $q$ under GND semantics. To efficiently tackle the S$^3$GND problem, we propose an effective learning-based approach, which constructs a keyword hypergraph from the data graph, and trains a \textit{hypergraph neural network} (HGNN) model to obtain high-quality keyword embedding representations. We design effective pruning strategies, \textit{keyword embedding MBR}, \textit{vertex-Level ND lower bound}, and \textit{graph-level GND lower bound pruning}, to rule out false alarms of candidate vertices/subgraphs, and devise a tree-based indexing mechanism to facilitate efficient S$^3$GND query answering. We develop an efficient S$^3$GND query-processing algorithm that traverses the index, applies pruning strategies, and returns actual S$^3$GND answers. Finally, we conduct extensive experiments to verify the effectiveness and efficiency of our proposed S$^3$GND approach over both real and synthetic graphs.

cs.DB↗
arXiv2026

Error Notebook-Guided, Training-Free Part Retrieval in 3D CAD Assemblies via Vision-Language Models

Effective specification-aware part retrieval within complex CAD assemblies is essential for automated engineering tasks. However, using LLMs/VLMs for this task is challenging: the CAD model metadata sequences often exceed token budgets, and fine-tuning high-performing proprietary models (e.g., GPT or Gemini) is unavailable. Therefore, we need a framework that delivers engineering value by handling long, non-natural-language CAD model metadata using VLMs, but without training. We propose a 2-stage framework with inference-time adaptation that combines corrected Error Notebooks with RAG to substantially improve VLM-based part retrieval reasoning. Each Error Notebook is built by correcting initial CoTs through reflective refinement, and then filtering each trajectory using our proposed grammar-constraint (GC) verifier to ensure structural well-formedness. The resulting notebook forms a high-quality repository of specification-CoT-answer triplets, from which RAG retrieves specification-relevant exemplars to condition the model's inference. We additionally contribute a CAD dataset with human preference annotations. Experiments with proprietary models (GPT-4o, Gemini, etc) show large gains, with GPT-4o (Omni) achieving up to +23.4 absolute accuracy points on the human-preference benchmark. The proposed GC verifier can further produce up to +4.5 accuracy points. Our approach also surpasses other training-free baselines (standard few-shot learning, self-consistency) and yields substantial improvements also for open-source VLMs (Qwen2-VL-2B-Instruct, Aya-Vision-8B). Under the cross-model GC setting, where the Error Notebook is constructed using GPT-4o (Omni), the 2B model inference achieves performance that comes within roughly 4 points of GPT-4o mini.

cs.AI↗
arXiv2026

MaS-VQA: A Mask-and-Select Framework for Knowledge-Based Visual Question Answering

Knowledge-based Visual Question Answering (KB-VQA) requires models to answer questions by integrating visual information with external knowledge. However, retrieved knowledge is often noisy, partially irrelevant, or misaligned with the visual content, while internal model knowledge is difficult to control and interpret. Naive aggregation of these sources limits reasoning effectiveness and reduces answer accuracy. To address this, we propose MaS-VQA, a selection-driven framework that tightly couples explicit knowledge filtering with implicit knowledge reasoning. MaS-VQA first retrieves candidate passages and applies a Mask-and-Select mechanism to jointly prune irrelevant image regions and weakly relevant knowledge fragments, producing compact, high-signal multimodal knowledge . This filtered knowledge then guides the activation of internal knowledge in a constrained semantic space, enabling complementary co-modeling of explicit and implicit knowledge for robust answer prediction. Experiments on Encyclopedic-VQA and InfoSeek demonstrate consistent performance gains across multiple MLLM backbones, and ablations verify that the selection mechanism effectively reduces noise and enhances knowledge utilization.

cs.CV↗
arXiv2026

Vista: Scene-Aware Optimization for Streaming Video Question Answering under Post-Hoc Queries

Streaming video question answering (Streaming Video QA) poses distinct challenges for multimodal large language models (MLLMs), as video frames arrive sequentially and user queries can be issued at arbitrary time points. Existing solutions relying on fixed-size memory or naive compression often suffer from context loss or memory overflow, limiting their effectiveness in long-form, real-time scenarios. We present Vista, a novel framework for scene-aware streaming video QA that enables efficient and scalable reasoning over continuous video streams. The innovation of Vista can be summarized in three aspects: (1) scene-aware segmentation, where Vista dynamically clusters incoming frames into temporally and visually coherent scene units; (2) scene-aware compression, where each scene is compressed into a compact token representation and stored in GPU memory for efficient index-based retrieval, while full-resolution frames are offloaded to CPU memory; and (3) scene-aware recall, where relevant scenes are selectively recalled and reintegrated into the model input upon receiving a query, enabling both efficiency and completeness. Vista is model-agnostic and integrates seamlessly with a variety of vision-language backbones, enabling long-context reasoning without compromising latency or memory efficiency. Extensive experiments on StreamingBench demonstrate that Vista achieves state-of-the-art performance, establishing a strong baseline for real-world streaming video understanding.

cs.CV↗
arXiv2026

QuantLRM: Quantization of Large Reasoning Models via Fine-Tuning Signals

Weight-only quantization is important for compressing Large Language Models (LLMs). Inspired by the spirit of classical magnitude pruning, we study whether the magnitude of weight updates during reasoning-incentivized fine-tuning can provide valuable signals for quantizing Large Reasoning Models (LRMs). We hypothesize that the smallest and largest weight updates during fine-tuning are more important than those of intermediate magnitude, a phenomenon we term "protecting both ends". Upon hypothesis validation, we introduce QuantLRM, which stands for weight quantization of LRMs via fine-tuning signals. We fit simple restricted quadratic functions on weight updates to protect both ends. By multiplying the average quadratic values with the count of zero weight updates of channels, we compute channel importance that is more effective than using activation or second-order information. We run QuantLRM to quantize various fine-tuned models (including supervised, direct preference optimization, and reinforcement learning fine-tuning) over four reasoning benchmarks (AIME-120, FOLIO, temporal sequences, and GPQA-Diamond) and empirically find that QuantLRM delivers a consistent improvement for LRMs quantization, with an average improvement of 6.55% on a reinforcement learning fine-tuned model. Also supporting non-fine-tuned LRMs, QuantLRM gathers effective signals via pseudo-fine-tuning, which greatly enhances its applicability.

cs.LG↗
arXiv2026

Triage: Hierarchical Visual Budgeting for Efficient Video Reasoning in Vision-Language Models

Vision-Language Models (VLMs) face significant computational challenges in video processing due to massive data redundancy, which creates prohibitively long token sequences. To address this, we introduce Triage, a training-free, plug-and-play framework that reframes video reasoning as a resource allocation problem via hierarchical visual budgeting. Its first stage, Frame-Level Budgeting, identifies keyframes by evaluating their visual dynamics and relevance, generating a strategic prior based on their importance scores. Guided by this prior, the second stage, Token-Level Budgeting, allocates tokens in two phases: it first secures high-relevance Core Tokens, followed by diverse Context Tokens selected with an efficient batched Maximal Marginal Relevance (MMR) algorithm. Extensive experiments demonstrate that Triage improves inference speed and reduces memory footprint, while maintaining or surpassing the performance of baselines and other methods on various video reasoning benchmarks.

cs.CV↗
arXiv2026

A census of star-formation and gas mass tracers in two lensed $z \sim 4$ dusty star-forming galaxies

We present new and archival Atacama Large Millimeter/submillimeter Array (ALMA) observations of two strongly lensed dusty star-forming galaxies (DSFGs) selected from the South Pole Telescope survey, SPT0418-47 $(z = 4.225)$ and SPT2147-50 $(z = 3.760)$. We study the [C II], CO(7-6), [C I](2-1), and, in SPT0418-47, $p$-H$_2$O emission, which along with the underlying continuum (rest-frame 160 $μ$m and 380 $μ$m) are routinely used as tracers of gas mass and/or star-formation rate (SFR). We perform a pixel-by-pixel analysis of both sources in the image plane to study the resolved Kennicutt-Schmidt relation, finding generally good agreement between the slopes of the SFR versus gas mass surface density using the different tracers. Using lens modeling methods, we find that the dust emission is more compact than the line emission in both sources, with CO(7-6) and [C I](2-1) similar in extent and [C II] the most extended, reminiscent of recent findings of extended [C II] spatial distributions in galaxies at similar cosmic epochs. We develop the [C I](2-1) / CO(7-6) flux density ratio as an observable proxy for gas depletion timescale ($τ_{\rm dep}$), which can be applied to large samples of DSFGs, in lieu of more detailed inferences of this timescale which require analysis of observations at multiple wavelengths. Furthermore, the extended [C II] emission in both sources, compared to the total continuum and line emission, suggests that [C II], used in recent years as a molecular gas mass and SFR tracer in high-$z$ galaxies, may not always be a suitable tracer of these physical quantities.

astro-ph.GA↗