Searcharxiv⌕ Search

arXiv subjects

Xiangwen Zhang

Publications and source records attributed to Xiangwen Zhang.

At least 19 recordsLinked to original sources
arXiv2026

AccidentSim: Generating Vehicle Collision Videos with Physically Realistic Collision Trajectories from Real-World Accident Reports

Collecting real-world vehicle accident videos for autonomous driving research is challenging due to their rarity and complexity. While existing driving video generation methods may produce visually realistic videos, they often fail to deliver physically realistic simulations because they lack the capability to generate accurate post-collision trajectories. In this paper, we introduce AccidentSim, a novel framework that generates physically realistic vehicle collision videos by extracting and utilizing the physical clues and contextual information available in real-world vehicle accident reports. Specifically, AccidentSim leverages a reliable physical simulator to replicate post-collision vehicle trajectories from the physical and contextual information in the accident reports and to build a vehicle collision trajectory dataset. This dataset is then used to fine-tune a language model, enabling it to respond to user prompts and predict physically consistent post-collision trajectories across various driving scenarios based on user descriptions. Finally, we employ Neural Radiance Fields (NeRF) to render high-quality backgrounds, merging them with the foreground vehicles that exhibit physically realistic trajectories to generate vehicle collision videos. Experimental results demonstrate that the videos produced by AccidentSim excel in both visual and physical authenticity.

cs.CV↗
arXiv2026

MapTab: A Diagnostic Benchmark for Long-Horizon Multi-Criteria Multimodal Reasoning on Heterogeneous Topological Graphs

Systematically evaluating Multimodal Large Language Models (MLLMs) is essential for advancing Artificial General Intelligence (AGI). Yet existing benchmarks remain inadequate for rigorously measuring their reasoning capabilities under multi-criteria constraints. To address this gap, we introduce MapTab, a multimodal benchmark designed to assess holistic multi-criteria reasoning in MLLMs through route-planning tasks. MapTab requires models to perceive and ground visual information from map images while integrating route attributes, such as Time and Price, from structured tables. It covers two scenarios: Metromap, spanning metro networks in 160 cities across 52 countries, and Travelmap, featuring 168 representative tourist attractions from 19 countries. Overall, MapTab includes 328 images, 196,800 route-planning queries, and 3,936 QA queries, incorporating four key criteria: Time, Price, Comfort, and Reliability. Extensive evaluations of 21 representative MLLMs show that current models still struggle with multicriteria multimodal reasoning. Notably, when visual perception is unreliable, multimodal reasoning can even underperform unimodal approaches. MapTab therefore offers a challenging and realistic testbed for systematically evaluating and advancing MLLMs across core perception, integration, numerical comparison, and route planning capabilities.

cs.LG↗
arXiv2026

LivePhys: Transforming Static Physics Problems into Interactive Simulations via a Scan-to-Play Framework

Physics problems in textbooks are typically presented as static diagrams accompanied by brief textual descriptions, requiring learners to infer dynamic physical behaviors through mental visualization. This process often imposes high cognitive demands and limits learners' ability to form accurate mental models. In this paper, we present \textbf{LivePhys}, a framework that enables a \emph{Scan-to-Play} paradigm for mechanics learning by transforming static textbook physics problems into executable, interactive simulations. LivePhys decouples multimodal perception from physics-aware reasoning and deterministic simulation. Given a problem diagram and its accompanying text, LivePhys performs text extraction, geometric segmentation, and cross-modal grounding to construct a structured, physics-aware intermediate representation. A multimodal large language model is then used as a reasoning controller to infer entities, parameters, and constraints, which are executed by a physics engine to generate spatially consistent and interactive simulations that allow learners to explore and manipulate problem conditions dynamically. Our evaluation results show that LivePhys significantly outperforms general-purpose multimodal models in simulation executability, spatial accuracy, and interaction fidelity. In addition, a user study demonstrates that interacting with LivePhys-generated simulations reduces learners' perceived cognitive load compared to static textbook materials.

cs.ET↗
arXiv2026

Multi-Stakeholder LLM Alignment: Decomposing Estimation from Aggregation

Multi-stakeholder tasks require one output to satisfy users with conflicting preferences. Holistic LLM judges conflate utility estimation and utility aggregation, yielding unstable implicit weights. We show empirically and theoretically that this aggregation-specific \emph{weighting noise} can create large score shifts when stakeholder satisfaction is dispersed; in our experiments, these weight-induced shifts also increase with stakeholder count. We propose \textsc{DecompR}: counterfactual-calibrated weights are fixed from query structure before candidate scoring, while per-role utilities are estimated independently, removing candidate-dependent weight drift and reducing estimation noise.

cs.AI↗
arXiv2026

Beyond Itinerary Planning-A Real-World Benchmark for Multi-Turn and Tool-Using Travel Tasks

Travel planning is a natural real-world task to test large language models' (LLMs) planning and tool-use abilities. Although prior work has studied LLM performance on travel planning, existing settings still differ from real-world needs, mainly due to limited domain coverage, insufficient modeling of users' implicit preferences in multi-turn conversations, and a lack of evaluation of agents' capability boundaries. To mitigate these gaps, we propose $\textbf{TravelBench}$, a benchmark for $\textit{truly real-world}$ travel planning. We collect user queries, user preferences, and tools from real scenarios, and construct three subtasks -- $\textit{Single-Turn}$, $\textit{Multi-Turn}$, and $\textit{Unsolvable}$ -- to evaluate agents' three core capabilities in real settings: (1) solving problems independently, (2) interacting with users to elicit implicit preferences, and (3) recognizing the capability boundaries. To enable stable tool invocation and reproducible evaluation, we cache real tool-call results and build a sandbox environment which integrates ten travel-related tools, enabling agents to combine these tools to solve most practical travel planning problems. We evaluate multiple LLMs on TravelBench and find that even advanced models exhibit imbalanced performance across different capabilities. Our further systematic verification demonstrates the stability of the proposed benchmark. TravelBench provides a practical and reproducible benchmark to advance research on LLM agents for real-world travel planning.

cs.AI↗
arXiv2026

No More Stale Feedback: Co-Evolving Critics for Open-World Agent Learning

Critique-guided reinforcement learning (RL) has emerged as a powerful paradigm for training LLM agents by augmenting sparse outcome rewards with natural-language feedback. However, current methods often rely on static or offline critic models, which fail to adapt as the policy evolves. In on-policy RL, the agent's error patterns shift over time, causing stationary critics to become stale and providing feedback of diminishing utility. To address this, we introduce ECHO (Evolving Critic for Hindsight-Guided Optimization)}, a framework that jointly optimizes the policy and critic through a synchronized co-evolutionary loop. ECHO utilizes a cascaded rollout mechanism where the critic generates multiple diagnoses for an initial trajectory, followed by policy refinement to enable group-structured advantage estimation. We address the challenge of learning plateaus via a saturation-aware gain shaping objective, which rewards the critic for inducing incremental improvements in high-performing trajectories. By employing dual-track GRPO updates, ECHO ensures the critic's feedback stays synchronized with the evolving policy. Experimental results show that ECHO yields more stable training and higher long-horizon task success across open-world environments.

cs.AI↗
arXiv2026

AMAP Agentic Planning Technical Report

We present STAgent, an agentic large language model tailored for spatio-temporal understanding, designed to solve complex tasks such as constrained point-of-interest discovery and itinerary planning. STAgent is a specialized model capable of interacting with ten distinct tools within spatio-temporal scenarios, enabling it to explore, verify, and refine intermediate steps during complex reasoning. Notably, STAgent effectively preserves its general capabilities. We empower STAgent with these capabilities through three key contributions: (1) a stable tool environment that supports over ten domain-specific tools, enabling asynchronous rollout and training; (2) a hierarchical data curation framework that identifies high-quality data like a needle in a haystack, curating high-quality queries by retaining less than 1\% of the raw data, emphasizing both diversity and difficulty; and (3) a cascaded training recipe that starts with a seed SFT stage acting as a guardian to measure query difficulty, followed by a second SFT stage fine-tuned on queries with high certainty, and an ultimate RL stage that leverages data of low certainty. Initialized with Qwen3-30B-A3B to establish a strong SFT foundation and leverage insights into sample difficulty, STAgent yields promising performance on TravelBench while maintaining its general capabilities across a wide range of general benchmarks, thereby demonstrating the effectiveness of our proposed agentic model.

cs.AI↗
arXiv2025

SketchPlay: Intuitive Creation of Physically Realistic VR Content with Gesture-Driven Sketching

Creating physically realistic content in VR often requires complex modeling tools or predefined 3D models, textures, and animations, which present significant barriers for non-expert users. In this paper, we propose SketchPlay, a novel VR interaction framework that transforms humans' air-drawn sketches and gestures into dynamic, physically realistic scenes, making content creation intuitive and playful like drawing. Specifically, sketches capture the structure and spatial arrangement of objects and scenes, while gestures convey physical cues such as velocity, direction, and force that define movement and behavior. By combining these complementary forms of input, SketchPlay captures both the structure and dynamics of user-created content, enabling the generation of a wide range of complex physical phenomena, such as rigid body motion, elastic deformation, and cloth dynamics. Experimental results demonstrate that, compared to traditional text-driven methods, SketchPlay offers significant advantages in expressiveness, and user experience. By providing an intuitive and engaging creation process, SketchPlay lowers the entry barrier for non-expert users and shows strong potential for applications in education, art, and immersive storytelling.

cs.HC↗
arXiv2025

Finedeep: Mitigating Sparse Activation in Dense LLMs via Multi-Layer Fine-Grained Experts

Large language models have demonstrated exceptional performance across a wide range of tasks. However, dense models usually suffer from sparse activation, where many activation values tend towards zero (i.e., being inactivated). We argue that this could restrict the efficient exploration of model representation space. To mitigate this issue, we propose Finedeep, a deep-layered fine-grained expert architecture for dense models. Our framework partitions the feed-forward neural network layers of traditional dense models into small experts, arranges them across multiple sub-layers. A novel routing mechanism is proposed to determine each expert's contribution. We conduct extensive experiments across various model sizes, demonstrating that our approach significantly outperforms traditional dense architectures in terms of perplexity and benchmark performance while maintaining a comparable number of parameters and floating-point operations. Moreover, we find that Finedeep achieves optimal results when balancing depth and width, specifically by adjusting the number of expert sub-layers and the number of experts per sub-layer. Empirical results confirm that Finedeep effectively alleviates sparse activation and efficiently utilizes representation capacity in dense models.

cs.CL↗
arXiv2022

MSP: Multi-Stage Prompting for Making Pre-trained Language Models Better Translators

Prompting has recently been shown as a promising approach for applying pre-trained language models to perform downstream tasks. We present Multi-Stage Prompting (MSP), a simple and automatic approach for leveraging pre-trained language models to translation tasks. To better mitigate the discrepancy between pre-training and translation, MSP divides the translation process via pre-trained language models into multiple separate stages: the encoding stage, the re-encoding stage, and the decoding stage. During each stage, we independently apply different continuous prompts for allowing pre-trained language models better shift to translation tasks. We conduct extensive experiments on three translation tasks. Experiments show that our method can significantly improve the translation performance of pre-trained language models.

cs.CL↗
arXiv2021

Stability of the Type IIA flow and its applications in symplectic geometry

In this paper the dynamical stability of the Type IIA flow with no source near its stationary points is established. These stationary points had been shown previously by the authors to be Ricci-flat Kähler metrics on Calabi-Yau 3-folds. The dynamical stability of the Type IIA flow is then applied to prove the stability under symplectic deformations of the Kähler property for Calabi-Yau 3-folds.

math.DG↗
arXiv2020

Bochner-Kodaira Formulas and the Type IIA Flow

A new derivation of the flow of metrics in the Type IIA flow is given. It is adapted to the formulation of the flow as a variant of a Laplacian flow, and it uses the projected Levi-Civita connection of the metrics themselves instead of their conformal rescalings.

math.DG↗
arXiv2020

Geometric Flows for the Type IIA String

A geometric flow on $6$-dimensional symplectic manifolds is introduced which is motivated by supersymmetric compactifications of the Type IIA string. The underlying structure turns out to be SU(3) holonomy, but with respect to the projected Levi-Civita connection of an almost-Hermitian structure. The short-time existence is established, and new identities for the Nijenhuis tensor are found which are crucial for Shi-type estimates. The integrable case can be completely solved, giving an alternative proof of Yau's theorem on Ricci-flat Kähler metrics. In the non-integrable case, models are worked out which suggest that the flow should lead to optimal almost-complex structures compatible with the given symplectic form.

math.DG↗
arXiv2020

Estimates for a geometric flow for the Type IIB string

It is shown that bounds of all orders of derivative would follow from uniform bounds for the metric and the torsion 1-form, for a flow in non-Kähler geometry which can be interpreted as either a flow for the Type IIB string or the Anomaly flow with source term and zero slope parameter. A key ingredient in the proof is a formulation of this flow unifying it with the Ricci flow, which was recently found.

math.DG↗
arXiv2019

A class of curvature type equations

In this paper, we study the solvability of a general class of fully nonlinear curvature equations, which can be viewed as generalizations of the equations for Christoffel-Minkowski problem in convex geometry. We will also study the Dirichlet problem of the corresponding degenerate equations as an extension of the equations studied by Krylov.

math.AP↗
arXiv2018

Otem&Utem: Over- and Under-Translation Evaluation Metric for NMT

Although neural machine translation(NMT) yields promising translation performance, it unfortunately suffers from over- and under-translation is- sues [Tu et al., 2016], of which studies have become research hotspots in NMT. At present, these studies mainly apply the dominant automatic evaluation metrics, such as BLEU, to evaluate the overall translation quality with respect to both adequacy and uency. However, they are unable to accurately measure the ability of NMT systems in dealing with the above-mentioned issues. In this paper, we propose two quantitative metrics, the Otem and Utem, to automatically evaluate the system perfor- mance in terms of over- and under-translation respectively. Both metrics are based on the proportion of mismatched n-grams between gold ref- erence and system translation. We evaluate both metrics by comparing their scores with human evaluations, where the values of Pearson Cor- relation Coefficient reveal their strong correlation. Moreover, in-depth analyses on various translation systems indicate some inconsistency be- tween BLEU and our proposed metrics, highlighting the necessity and significance of our metrics.

cs.CL↗
arXiv2018

New curvature flows in complex geometry

This is a survey of some of the recent developments on the geometric and analytic aspects of the Anomaly flow. It is a flow of $(2,2)$-forms on a $3$-fold which was originally motivated by string theory and the need to preserve the conformally balanced property of a Hermitian metric in the absence of a $\partial\bar\partial$-Lemma. It has revealed itself since to be a remarkable higher order extension of the Ricci flow. It has also led to several other curvature flows which may be interesting from the point of view of both non-Kähler geometry and the theory of non-linear partial differential equations.

math.DG↗