arXiv subjects

Mao Zheng

Publications and source records attributed to Mao Zheng.

44 records · Page 3Linked to original sources

arXiv2024

MiMoTable: A Multi-scale Spreadsheet Benchmark with Meta Operations for Table Reasoning

Extensive research has been conducted to explore the capability of Large Language Models (LLMs) for table reasoning and has significantly improved the performance on existing benchmarks. However, tables and user questions in real-world applications are more complex and diverse, presenting an unignorable gap compared to the existing benchmarks. To fill the gap, we propose a \textbf{M}ult\textbf{i}-scale spreadsheet benchmark with \textbf{M}eta \textbf{o}perations for \textbf{Table} reasoning, named as MiMoTable. Specifically, MiMoTable incorporates two key features. First, the tables in MiMoTable are all spreadsheets used in real-world scenarios, which cover seven domains and contain different types. Second, we define a new criterion with six categories of meta operations for measuring the difficulty of each question in MiMoTable, simultaneously as a new perspective for measuring the difficulty of the existing benchmarks. Experimental results show that Claude-3.5-Sonnet achieves the best performance with 77.4\% accuracy, indicating that there is still significant room to improve for LLMs on MiMoTable. Furthermore, we grade the difficulty of existing benchmarks according to our new criteria. Experiments have shown that the performance of LLMs decreases as the difficulty of benchmarks increases, thereby proving the effectiveness of our proposed new criterion.

cs.CL↗

arXiv2023

STOA-VLP: Spatial-Temporal Modeling of Object and Action for Video-Language Pre-training

Although large-scale video-language pre-training models, which usually build a global alignment between the video and the text, have achieved remarkable progress on various downstream tasks, the idea of adopting fine-grained information during the pre-training stage is not well explored. In this work, we propose STOA-VLP, a pre-training framework that jointly models object and action information across spatial and temporal dimensions. More specifically, the model regards object trajectories across frames and multiple action features from the video as fine-grained features. Besides, We design two auxiliary tasks to better incorporate both kinds of information into the pre-training process of the video-language model. The first is the dynamic object-text alignment task, which builds a better connection between object trajectories and the relevant noun tokens. The second is the spatial-temporal action set prediction, which guides the model to generate consistent action features by predicting actions found in the text. Extensive experiments on three downstream tasks (video captioning, text-video retrieval, and video question answering) demonstrate the effectiveness of our proposed STOA-VLP (e.g. 3.7 Rouge-L improvements on MSR-VTT video captioning benchmark, 2.9% accuracy improvements on MSVD video question answering benchmark, compared to previous approaches).

cs.CV↗

arXiv2022

Alignment-Uniformity aware Representation Learning for Zero-shot Video Classification

Most methods tackle zero-shot video classification by aligning visual-semantic representations within seen classes, which limits generalization to unseen classes. To enhance model generalizability, this paper presents an end-to-end framework that preserves alignment and uniformity properties for representations on both seen and unseen classes. Specifically, we formulate a supervised contrastive loss to simultaneously align visual-semantic features (i.e., alignment) and encourage the learned features to distribute uniformly (i.e., uniformity). Unlike existing methods that only consider the alignment, we propose uniformity to preserve maximal-info of existing features, which improves the probability that unobserved features fall around observed data. Further, we synthesize features of unseen classes by proposing a class generator that interpolates and extrapolates the features of seen classes. Besides, we introduce two metrics, closeness and dispersion, to quantify the two properties and serve as new measurements of model generalizability. Experiments show that our method significantly outperforms SoTA by relative improvements of 28.1% on UCF101 and 27.0% on HMDB51. Code is available.

cs.CV↗

arXiv2020

Multimodal Topic Learning for Video Recommendation

Facilitated by deep neural networks, video recommendation systems have made significant advances. Existing video recommendation systems directly exploit features from different modalities (e.g., user personal data, user behavior data, video titles, video tags, and visual contents) to input deep neural networks, while expecting the networks to online mine user-preferred topics implicitly from these features. However, the features lacking semantic topic information limits accurate recommendation generation. In addition, feature crosses using visual content features generate high dimensionality features that heavily downgrade the online computational efficiency of networks. In this paper, we explicitly separate topic generation from recommendation generation, propose a multimodal topic learning algorithm to exploit three modalities (i.e., tags, titles, and cover images) for generating video topics offline. The topics generated by the proposed algorithm serve as semantic topic features to facilitate preference scope determination and recommendation generation. Furthermore, we use the semantic topic features instead of visual content features to effectively reduce online computational cost. Our proposed algorithm has been deployed in the Kuaibao information streaming platform. Online and offline evaluation results show that our proposed algorithm performs favorably.

cs.IR↗

arXiv2016

Modeling The Adaption Rule in Context-aware Systems

Context awareness is increasingly gaining applicability in interactive ubiquitous mobile computing systems. Each context-aware application has its own set of behaviors to react to context modifications. This paper is concerned with the context modeling and the development methodology for context-aware systems. We proposed a rule-based approach and use the adaption tree to model the adaption rule of context-aware systems. We illustrate this idea in an arithmetic game application.

cs.HC↗

arXiv2013

Uniform Oxygen Doping Leads to Superconductivity in FeTe Films

FeTe is known to become a superconductor when doped with oxygen. Using layer by layer growth of single crystal films by molecular beam epitaxy (MBE), we have studied how oxygen incorporates. If oxygen is supplied during growth of a layer, it substitutes for tellurium inhomogeneously in oxygen domains that are not associated with superconductivity. When oxygen is supplied after growth, it diffuses homogeneously into the crystalline film and incorporates interstitially. Only the interstitial oxygen causes superconductivity to emerge. This suggests that the superconductivity observed in this material is spatially uniform and not filamentary.

cond-mat.supr-con↗

arXiv2011

Growth and Properties of Heavy Fermion CeCu2Ge2 and CeFe2Ge2 thin films

Epitaxial films of heavy fermion CeCu2Ge2 and CeFe2Ge2 are grown on DyScO3 and MgO substrates using molecular beam epitaxy. The growth begins via island nucleation leading to a granular morphology. The grains grow flat with c-axis orientation after nucleating, as indicated by in-situ reflection high energy electron diffraction (RHEED) and ex-situ analysis including atomic force microscopy (AFM) and x-ray diffraction (XRD). These single phase films show similar temperature dependent transport to single crystals of the materials indicating that similar collective order occurs in the films as in single crystals.

cond-mat.mtrl-sci↗

arXiv2009

Radiative lifetime measurements of rubidium Rydberg states

We have measured the radiative lifetimes of ns, np and nd Rydberg states of rubidium in the range 28 < n < 45. To enable long-lived states to be measured, our experiment uses slow-moving Rb atoms in a magneto-optical trap (MOT). Two experimental techniques have been adopted to reduce random and systematic errors. First, a narrow-bandwidth pulsed laser is used to excite the target Rydberg state, resulting in minimal shot-to-shot variation in the initial state population. Second, we monitor the target state population as a function of time delay from the laser pulse using a short-duration, millimetre-wave pulse that is resonant with a one- or two-photon transition. We then selectively field ionize the monitor state, and detect the resulting electrons with a micro-channel plate. This signal is an accurate mirror of the target state population, and is uncontaminated by contributions from other states which are populated by black body radiation. Our results are generally consistent with other recent experimental results obtained using a less sensitive method, and are also in excellent agreement with theory.

physics.atom-ph↗