arXiv subjects

Yuanzhe Wang

Publications and source records attributed to Yuanzhe Wang.

9 recordsLinked to original sources

arXiv2026

XL-DocBench: Benchmarking Evidence-Grounded Extra-Long Document Understanding

Real-world document tasks often ask professionals to answer questions from annual reports, regulations, clinical guidelines, and technical manuals that span hundreds or thousands of pages. Some questions also require comparing related reports. Reliable long-document understanding is therefore a prerequisite for using LLMs in compliance, clinical, financial, and engineering workflows, where decisions must be traceable to specific evidence pages and the cost of an unsupported answer is high -- yet most existing benchmarks still measure short-context or single-page QA. We introduce XL-DocBench, a fully human-verified benchmark for extra-long document understanding, with 1,519 retained questions from six professional domains and contexts up to 2,303 pages. XL-DocBench goes beyond page-level lookup. 1,103 examples (72.6\%) use multiple evidence pages. The final set also includes 556 questions (36.6\%) that use tables, charts, or figures, and 165 questions (10.9\%) that require evidence from multiple documents. Each question has one of twelve reasoning labels, expert-annotated evidence pages, a typed verification rule, and an answer format, including 218 None-answer cases. We build the benchmark with a tree-guided synthesis pipeline followed by artifact filters and full verification by 194 human experts. By coupling extra-long professional contexts with page-level evidence and typed rules, XL-DocBench fills a gap left by prior single-page, short multi-page, or text-only long-context benchmarks, and lets future work attribute system failures to retrieval, evidence use, or rule following rather than to a single leaderboard score. The results show that current systems still struggle with long contexts, multi-page evidence, and structured reasoning over professional documents.

cs.CL↗

arXiv2026

DocAtlas: Long-Document Understanding as Mutable-State Interaction

Long-document understanding requires models to find and combine evidence across many pages, layouts, tables, figures, and charts. Existing retrieval-augmented systems usually select evidence from a static index before generation, while recent agentic systems add multi-turn tool use but often rely on frozen proprietary backbones whose behavior is set by prompts. We present DocAtlas, a system that treats long-document understanding as a mutable-state information-seeking process. We instantiate DocAtlas as a mutable document harness: an external environment that determines what document information is searched, read, stored, reviewed, and shown to the model at each step. Given a document and question, the harness exposes search, reading, note-taking, and review tools, maintains a hierarchical tree and note store, and updates both as the agent records evidence. DocAtlas combines self-improving retrieval, selective evidence access, and active working memory under a fixed context budget. The same harness supports inference-time use with large VLMs and end-to-end reinforcement learning for compact VLM agents. With GPT-5.4, DocAtlas reaches 71.4\% on MMLongBench-Doc, exceeding the human-expert reference of 65.8\%. A Qwen3.5-4B VLM trained with end-to-end RL in the DocAtlas environment reaches 63.7\%, compared with a 54.4\% direct-input baseline, showing that mutable document-harness design can improve compact document agents by a large margin.

cs.CL↗

arXiv2026

Latent Diffusion Posterior Sampling with Surrogate Likelihood Guidance for PDE Inverse Problems

We propose latent-space diffusion posterior sampling (L-DPS), an approximate Bayesian framework for high-dimensional inverse problems governed by partial differential equations (PDEs). The method addresses three challenges in PDE-constrained inversion: implicit sample-based priors without tractable densities, high-dimensional spatially distributed parameters, and the high cost of repeated forward-model evaluations during posterior sampling. L-DPS combines a variational autoencoder, an unconditional latent diffusion model, diffusion posterior sampling, and a differentiable neural surrogate. The VAE maps the parameter field to a lower-dimensional latent space, the diffusion model learns an implicit prior score in this latent space, and DPS combines this learned prior with likelihood-based guidance. The likelihood gradient is evaluated through the decoder-surrogate composition, avoiding repeated calls to the full numerical PDE solver. We evaluate the method on an inverse Darcy flow problem with an unknown spatially distributed permeability field inferred from sparse and noisy pressure observations. L-DPS produces accurate and robust inverse solutions, reduces inference cost relative to full-space DPS, and outperforms amortized inverse baselines such as conditional latent diffusion and inverse FNO in sparse and noisy regimes. We further compare L-DPS with a KLE-MAP baseline and study mixed-prior generalization and the sensitivity of inversion accuracy to surrogate forward-model error.

cs.CE↗

arXiv2026

Discrete Diffusion for Complex and Congested Multi-Agent Path Finding with Sparse Social Attention

Multi-Agent Path Finding (MAPF) is a coordination problem that requires computing globally consistent, collision-free trajectories from individual start positions to assigned goal positions under combinatorial planning complexity. In dense environments, suboptimal initial plans induce compound conflicts that hinder feasible repair. For repair-based solvers like LNS2, initial plan quality critically affects downstream repair, yet this factor remains underexplored. We propose DiffLNS, a hybrid framework that integrates a discrete denoising diffusion probabilistic model (D3PM) with LNS2. The D3PM serves as an initializer with sparse social attention that learns a spatiotemporal prior over coordinated multi-agent action trajectories from expert demonstrations and samples multiple joint plans. Operating directly on the categorical action space, our discrete diffusion preserves the MAPF action structure and samples from a multimodal joint-plan distribution to produce diverse drafts well suited for neighborhood repair. These drafts act as warm starts for downstream repair, which completes unfinished trajectories and resolves remaining conflicts under hard MAPF constraints. Experimental results show that despite being trained only on instances with at most 96 agents, the initializer generalizes to scenarios with up to 312 agents at inference time. Across 20 complex and congested settings, DiffLNS achieves an average success rate of 95.8%, outperforming the strongest tested baseline by 9.6 percentage points and matching or exceeding all baselines in all 20 settings. To the best of our knowledge, this is the first work to leverage discrete diffusion for warm-starting an LNS-based MAPF solver.

cs.AI↗

arXiv2025

Low-Complexity Cooperative Payload Transportation for Nonholonomic Mobile Robots Under Scalable Constraints

Cooperative transportation, a key aspect of logistics cyber-physical systems (CPS), is typically approached using dis tributed control and optimization-based methods. The distributed control methods consume less time, but poorly handle and extend to multiple constraints. Instead, optimization-based methods handle constraints effectively, but they are usually centralized, time-consuming and thus not easily scalable to numerous robots. To overcome drawbacks of both, we propose a novel cooperative transportation method for nonholonomic mobile robots by im proving conventional formation control, which is distributed, has a low time-complexity and accommodates scalable constraints. The proposed control-based method is testified on a cable suspended payload and divided into two parts, including robot trajectory generation and trajectory tracking. Unlike most time consuming trajectory generation methods, ours can generate trajectories with only constant time-complexity, needless of global maps. As for trajectory tracking, our control-based method not only scales easily to multiple constraints as those optimization based methods, but reduces their time-complexity from poly nomial to linear. Simulations and experiments can verify the feasibility of our method.

cs.RO↗

arXiv2024

Solving High-dimensional Inverse Problems Using Amortized Likelihood-free Inference with Noisy and Incomplete Data

We present a likelihood-free probabilistic inversion method based on normalizing flows for high-dimensional inverse problems. The proposed method is composed of two complementary networks: a summary network for data compression and an inference network for parameter estimation. The summary network encodes raw observations into a fixed-size vector of summary features, while the inference network generates samples of the approximate posterior distribution of the model parameters based on these summary features. The posterior samples are produced in a deep generative fashion by sampling from a latent Gaussian distribution and passing these samples through an invertible transformation. We construct this invertible transformation by sequentially alternating conditional invertible neural network and conditional neural spline flow layers. The summary and inference networks are trained simultaneously. We apply the proposed method to an inversion problem in groundwater hydrology to estimate the posterior distribution of the log-conductivity field conditioned on spatially sparse time-series observations of the system's hydraulic head responses.The conductivity field is represented with 706 degrees of freedom in the considered problem.The comparison with the likelihood-based iterative ensemble smoother PEST-IES method demonstrates that the proposed method accurately estimates the parameter posterior distribution and the observations' predictive posterior distribution at a fraction of the inference time of PEST-IES.

cs.LG↗

arXiv2024

Total Uncertainty Quantification in Inverse PDE Solutions Obtained with Reduced-Order Deep Learning Surrogate Models

We propose an approximate Bayesian method for quantifying the total uncertainty in inverse PDE solutions obtained with machine learning surrogate models, including operator learning models. The proposed method accounts for uncertainty in the observations and PDE and surrogate models. First, we use the surrogate model to formulate a minimization problem in the reduced space for the maximum a posteriori (MAP) inverse solution. Then, we randomize the MAP objective function and obtain samples of the posterior distribution by minimizing different realizations of the objective function. We test the proposed framework by comparing it with the iterative ensemble smoother and deep ensembling methods for a non-linear diffusion equation with an unknown space-dependent diffusion coefficient. Among other problems, this equation describes groundwater flow in an unconfined aquifer. Depending on the training dataset and ensemble sizes, the proposed method provides similar or more descriptive posteriors of the parameters and states than the iterative ensemble smoother method. Deep ensembling underestimates uncertainty and provides less informative posteriors than the other two methods.

cs.LG↗

arXiv2023

NTU4DRadLM: 4D Radar-centric Multi-Modal Dataset for Localization and Mapping

Simultaneous Localization and Mapping (SLAM) is moving towards a robust perception age. However, LiDAR- and visual- SLAM may easily fail in adverse conditions (rain, snow, smoke and fog, etc.). In comparison, SLAM based on 4D Radar, thermal camera and IMU can work robustly. But only a few literature can be found. A major reason is the lack of related datasets, which seriously hinders the research. Even though some datasets are proposed based on 4D radar in past four years, they are mainly designed for object detection, rather than SLAM. Furthermore, they normally do not include thermal camera. Therefore, in this paper, NTU4DRadLM is presented to meet this requirement. The main characteristics are: 1) It is the only dataset that simultaneously includes all 6 sensors: 4D radar, thermal camera, IMU, 3D LiDAR, visual camera and RTK GPS. 2) Specifically designed for SLAM tasks, which provides fine-tuned ground truth odometry and intentionally formulated loop closures. 3) Considered both low-speed robot platform and fast-speed unmanned vehicle platform. 4) Covered structured, unstructured and semi-structured environments. 5) Considered both middle- and large- scale outdoor environments, i.e., the 6 trajectories range from 246m to 6.95km. 6) Comprehensively evaluated three types of SLAM algorithms. Totally, the dataset is around 17.6km, 85mins, 50GB and it will be accessible from this link: https://github.com/junzhang2016/NTU4DRadLM

cs.RO↗

arXiv2022

An SBR Based Ray Tracing Channel Modeling Method for THz and Massive MIMO Communications

Terahertz (THz) communication and the application of massive multiple-input multiple-output (MIMO) technology have been proved significant for the sixth generation (6G) communication systems, and have gained global interests. In this paper, we employ the shooting and bouncing ray (SBR) method integrated with acceleration technology to model THz and massive MIMO channel. The results of ray tracing (RT) simulation in this paper, i.e., angle of departure (AoD), angle of arrival (AoA), and power delay profile (PDP) under the frequency band supported by the commercial RT software Wireless Insite (WI) are in agreement with those produced by WI. Based on the Kirchhoff scattering effect on material surfaces and atmospheric absorption loss showing at THz frequency band, the modified propagation models of Fresnel reflection coefficients and free-space attenuation are consistent with the measured results. For massive MIMO, the channel capacity and the stochastic power distribution are analyzed. The results indicate the applicability of SBR method for building deterministic models of THz and massive MIMO channels with extensive functions and acceptable accuracy.

eess.SP↗