arXiv subjects

Yaqi Wu

Publications and source records attributed to Yaqi Wu.

9 recordsLinked to original sources

arXiv2026

DynaKRAG: A Unified Framework for Learnable Evidence Control in Multi-Hop Retrieval-Augmented Generation

Multi-hop retrieval-augmented generation (RAG) acquires evidence sequentially, with each document contributing supporting facts, bridge entities, query refinements, or sufficient evidence for answering. Evidence acquisition can involve iterative retrieval, query reformulation, evidence assessment, and sufficiency checking. We introduce DynaKRAG, a unified evidence-action framework that learns a shared state-conditioned policy for coordinating these operations. At each step, a deterministic validity layer constructs the executable action set, a learned continuation gate selects between answer generation and further evidence acquisition, and a learned advantage scorer ranks feasible evidence operations by their predicted gain relative to immediate answer generation. The selected operation updates the shared state and may enable additional operations. Across HotpotQA, 2Wiki, and MuSiQue with Qwen2.5-7B, GPT-4o-mini, and Llama-3.1-8B, DynaKRAG ranks first among the compared methods in both EM and F1 for all nine dataset--backbone pairs. Relative to matched-backbone baseline method, DynaKRAG improves F1 in every pair while achieving total-token efficiency gains of 10.1--34.3\% and retrieval-call efficiency gains of 15.1--43.4\%, establishing Pareto dominance under these measures. With Qwen2.5-7B, terminal evidence compression further improves answer quality across all three datasets while reducing the context passed to final answer generation by 54.4\%--71.5\%. These results demonstrate that unified, state-conditioned evidence control supports strong answer quality, efficient retrieval, and compact answer-generation contexts.

cs.CL↗

arXiv2026

RAFT-UP: Robust Alignment for Spatial Transcriptomics with Explicit Control of Spatial Distortion

Spatial transcriptomics (ST) profiles gene expression across a tissue section while preserving the spatial coordinates. Because current ST technologies typically profile two-dimensional tissue slices, integrating and aligning slices from different regions of the same three-dimensional tissue or from samples under different conditions enables analyses that reveal 3D organization and condition-associated spatial patterns. Two major challenges remain. First, interpretable and flexible control over spatial distortion is needed because rigid transformations can be overly restrictive, whereas highly deformable mappings may arbitrarily distort spatial proximity. Second, biologically plausible matching is also needed, especially when the slices overlap partially. Here, we introduce RAFT-UP, a tool for robust ST alignment that provides explicit control over spatial distance preservation through a fused supervised Gromov-Wasserstein (FsGW) optimal transport framework. FsGW combines expression and spatial information, incorporates spot-wise constraints to discourage biologically implausible matches, and enforces a pairwise distance-consistency constraint that prevents mapping two pairs of spots when their spatial distances differ beyond a specified tolerance. We demonstrate that RAFT-UP accurately aligns slices from different regions of the same tissue and slices from different samples. Benchmarking shows that RAFT-UP improves spatial distance preservation while achieving spot label matching accuracy comparable to state-of-the-art methods. Finally, we demonstrate RAFT-UP on two spatially constrained downstream applications, including spatiotemporal mapping of developing mouse midbrain and comparative cross-slice analysis of cell-cell communication. RAFT-UP is available as open-source software.

q-bio.QM↗

arXiv2025

OTMol: Robust Molecular Structure Comparison via Optimal Transport

Root-mean-square deviation (RMSD) is widely used to assess structural similarity in systems ranging from flexible ligand conformers to complex molecular cluster configurations. Despite its wide utility, RMSD calculation is often challenged by inconsistent atom ordering, indistinguishable configurations in molecular clusters, and potential chirality inversion during alignment. These issues highlight the necessity of accurate atom-to-atom correspondence as a prerequisite for meaningful alignment. Traditional approaches often rely on heuristic cost matrices combined with the Hungarian algorithm, yet these methods underutilize the rich intra-molecular structural information and may fail to generalize across chemically diverse systems. In this work, we introduce OTMol, a method that formulates the molecular alignment task as a fused supervised Gromov-Wasserstein (fsGW) optimal transport problem. By leveraging the intrinsic geometric and topological relationships within each molecule, OTMol eliminates the need for manually defined cost functions and enables a principled, data-driven matching strategy. Importantly, OTMol preserves key chemical features such as molecular chirality and bond connectivity consistency. We evaluate OTMol across a wide range of molecular systems, including Adenosine triphosphate, Imatinib, lipids, small peptides, and water clusters, and demonstrate that it consistently achieves low RMSD values while preserving computational efficiency. Importantly, OTMol maintains molecular integrity by enforcing one-to-one mappings between entire molecules, thereby avoiding erroneous many-to-one alignments that often arise in comparing molecular clusters. Our results underscore the utility of optimal transport theory for molecular alignment and offer a generalizable framework applicable to structural comparison tasks in cheminformatics, molecular modeling, and related disciplines.

q-bio.BM↗

arXiv2024

MIPI 2024 Challenge on Demosaic for HybridEVS Camera: Methods and Results

The increasing demand for computational photography and imaging on mobile platforms has led to the widespread development and integration of advanced image sensors with novel algorithms in camera systems. However, the scarcity of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). Building on the achievements of the previous MIPI Workshops held at ECCV 2022 and CVPR 2023, we introduce our third MIPI challenge including three tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Nighttime Flare Removal track on MIPI 2024. In total, 170 participants were successfully registered, and 14 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Nighttime Flare Removal. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2024/.

eess.IV↗

arXiv2024

TD^2-Net: Toward Denoising and Debiasing for Dynamic Scene Graph Generation

Dynamic scene graph generation (SGG) focuses on detecting objects in a video and determining their pairwise relationships. Existing dynamic SGG methods usually suffer from several issues, including 1) Contextual noise, as some frames might contain occluded and blurred objects. 2) Label bias, primarily due to the high imbalance between a few positive relationship samples and numerous negative ones. Additionally, the distribution of relationships exhibits a long-tailed pattern. To address the above problems, in this paper, we introduce a network named TD$^2$-Net that aims at denoising and debiasing for dynamic SGG. Specifically, we first propose a denoising spatio-temporal transformer module that enhances object representation with robust contextual information. This is achieved by designing a differentiable Top-K object selector that utilizes the gumbel-softmax sampling strategy to select the relevant neighborhood for each object. Second, we introduce an asymmetrical reweighting loss to relieve the issue of label bias. This loss function integrates asymmetry focusing factors and the volume of samples to adjust the weights assigned to individual samples. Systematic experimental results demonstrate the superiority of our proposed TD$^2$-Net over existing state-of-the-art approaches on Action Genome databases. In more detail, TD$^2$-Net outperforms the second-best competitors by 12.7 \% on mean-Recall@10 for predicate classification.

cs.CV↗

arXiv2024

Supervised Gromov-Wasserstein Optimal Transport

We introduce the supervised Gromov-Wasserstein (sGW) optimal transport, an extension of Gromov-Wasserstein by incorporating potential infinity patterns in the cost tensor. sGW enables the enforcement of application-induced constraints such as the preservation of pairwise distances by implementing the constraints as an infinity pattern. A numerical solver is proposed for the sGW problem and the effectiveness is demonstrated in various numerical experiments. The high-order constraints in sGW are transferred to constraints on the coupling matrix by solving a minimal vertex cover problem. The transformed problem is solved by the Mirror-C descent iteration coupled with the supervised optimal transport solver. In the numerical experiments, we first validate the proposed framework by applying it to matching synthetic datasets and investigating the impact of the model parameters. Additionally, we successfully apply sGW to real single-cell RNA sequencing data. Through comparisons with other Gromov-Wasserstein variants on real data, we demonstrate that sGW offers the novel utility of controlling distance preservation, leading to the automatic estimation of overlapping portions of datasets, which brings improved stability and flexibility in data-driven applications.

math.OC↗

arXiv2022

Learned Smartphone ISP on Mobile GPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report

The role of mobile cameras increased dramatically over the past few years, leading to more and more research in automatic image quality enhancement and RAW photo processing. In this Mobile AI challenge, the target was to develop an efficient end-to-end AI-based image signal processing (ISP) pipeline replacing the standard mobile ISPs that can run on modern smartphone GPUs using TensorFlow Lite. The participants were provided with a large-scale Fujifilm UltraISP dataset consisting of thousands of paired photos captured with a normal mobile camera sensor and a professional 102MP medium-format FujiFilm GFX100 camera. The runtime of the resulting models was evaluated on the Snapdragon's 8 Gen 1 GPU that provides excellent acceleration results for the majority of common deep learning ops. The proposed solutions are compatible with all recent mobile GPUs, being able to process Full HD photos in less than 20-50 milliseconds while achieving high fidelity results. A detailed description of all models developed in this challenge is provided in this paper.

cs.CV↗

arXiv2022

Efficient and Accurate Quantized Image Super-Resolution on Mobile NPUs, Mobile AI & AIM 2022 challenge: Report

Image super-resolution is a common task on mobile and IoT devices, where one often needs to upscale and enhance low-resolution images and video frames. While numerous solutions have been proposed for this problem in the past, they are usually not compatible with low-power mobile NPUs having many computational and memory constraints. In this Mobile AI challenge, we address this problem and propose the participants to design an efficient quantized image super-resolution solution that can demonstrate a real-time performance on mobile NPUs. The participants were provided with the DIV2K dataset and trained INT8 models to do a high-quality 3X image upscaling. The runtime of all models was evaluated on the Synaptics VS680 Smart Home board with a dedicated edge NPU capable of accelerating quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 60 FPS rate when reconstructing Full HD resolution images. A detailed description of all models developed in the challenge is provided in this paper.

eess.IV↗

arXiv2020

NTIRE 2020 Challenge on Real Image Denoising: Dataset, Methods and Results

This paper reviews the NTIRE 2020 challenge on real image denoising with focus on the newly introduced dataset, the proposed methods and their results. The challenge is a new version of the previous NTIRE 2019 challenge on real image denoising that was based on the SIDD benchmark. This challenge is based on a newly collected validation and testing image datasets, and hence, named SIDD+. This challenge has two tracks for quantitatively evaluating image denoising performance in (1) the Bayer-pattern rawRGB and (2) the standard RGB (sRGB) color spaces. Each track ~250 registered participants. A total of 22 teams, proposing 24 methods, competed in the final phase of the challenge. The proposed methods by the participating teams represent the current state-of-the-art performance in image denoising targeting real noisy images. The newly collected SIDD+ datasets are publicly available at: https://bit.ly/siddplus_data.

cs.CV↗