arXiv subjects

Haijing Liu

Publications and source records attributed to Haijing Liu.

11 recordsLinked to original sources

arXiv2026

Category-Adaptive Cross-Modal Semantic Refinement and Transfer for Open-Vocabulary Multi-Label Recognition

Benefiting from the generalization capability of CLIP, recent vision language pre-training (VLP) models have demonstrated the ability to capture a wide range of visual concepts in daily images. However, due to the presence of unseen categories in open-vocabulary settings, existing algorithms struggle to capture semantic correlations between categories, leading to suboptimal performance on open-vocabulary multi-label recognition (OV-MLR). Furthermore, the substantial variation in the number of discriminative areas across diverse object categories is misaligned with the fixed-number patch matching used in current methods, introducing noisy visual cues that hinder the capture of target semantics. To address these challenges, we propose a novel category-adaptive cross-modal semantic refinement and transfer (C$^2$SRT) framework to model semantic correlations both within each category and across different categories, in a category-adaptive manner. The proposed framework consists of two complementary modules, i.e., intra-category semantic refinement (ISR) module and inter-category semantic transfer (IST) module. Specifically, the ISR module leverages the cross-modal knowledge of the VLP model to adaptively select a set of local discriminative regions that represent the semantics of the target category. The IST module adaptively discovers a set of correlated categories for a target category by constructing a category-adaptive correlation graph and transfers semantic knowledge from the correlated seen categories to unseen ones. Experiments on OV-MLR benchmarks demonstrate that the proposed C$^2$SRT framework improves over current methods.

cs.CV↗

arXiv2025

Robust Egocentric Referring Video Object Segmentation via Dual-Modal Causal Intervention

Egocentric Referring Video Object Segmentation (Ego-RVOS) aims to segment the specific object actively involved in a human action, as described by a language query, within first-person videos. This task is critical for understanding egocentric human behavior. However, achieving such segmentation robustly is challenging due to ambiguities inherent in egocentric videos and biases present in training data. Consequently, existing methods often struggle, learning spurious correlations from skewed object-action pairings in datasets and fundamental visual confounding factors of the egocentric perspective, such as rapid motion and frequent occlusions. To address these limitations, we introduce Causal Ego-REferring Segmentation (CERES), a plug-in causal framework that adapts strong, pre-trained RVOS backbones to the egocentric domain. CERES implements dual-modal causal intervention: applying backdoor adjustment principles to counteract language representation biases learned from dataset statistics, and leveraging front-door adjustment concepts to address visual confounding by intelligently integrating semantic visual features with geometric depth information guided by causal principles, creating representations more robust to egocentric distortions. Extensive experiments demonstrate that CERES achieves state-of-the-art performance on Ego-RVOS benchmarks, highlighting the potential of applying causal reasoning to build more reliable models for broader egocentric video understanding.

cs.CV↗

arXiv2025

AndesVL Technical Report: An Efficient Mobile-side Multimodal Large Language Model

In recent years, while cloud-based MLLMs such as QwenVL, InternVL, GPT-4o, Gemini, and Claude Sonnet have demonstrated outstanding performance with enormous model sizes reaching hundreds of billions of parameters, they significantly surpass the limitations in memory, power consumption, and computing capacity of edge devices such as mobile phones. This paper introduces AndesVL, a suite of mobile-side MLLMs with 0.6B to 4B parameters based on Qwen3's LLM and various visual encoders. We comprehensively outline the model architectures, training pipeline, and training data of AndesVL, which achieves first-tier performance across a wide range of open-source benchmarks, including fields such as text-rich image understanding, reasoning and math, multi-image comprehension, general VQA, hallucination mitigation, multilingual understanding, and GUI-related tasks when compared with state-of-the-art models of a similar scale. Furthermore, we introduce a 1+N LoRA architecture alongside a Quantization-Aware LoRA Fine-Tuning (QALFT) framework to facilitate efficient task adaptation and model compression during mobile-side deployment of AndesVL. Moreover, utilizing our cache eviction algorithm -- OKV -- along with customized speculative decoding and compression strategies, we achieve a 6.7x peak decoding speedup ratio, up to 30.9% memory reduction, and 1.8 bits-per-weight when deploying AndesVL-4B on MediaTek Dimensity 9500 chips. We release all models on https://huggingface.co/OPPOer.

cs.CV↗

arXiv2025

DART: Dual Adaptive Refinement Transfer for Open-Vocabulary Multi-Label Recognition

Open-Vocabulary Multi-Label Recognition (OV-MLR) aims to identify multiple seen and unseen object categories within an image, requiring both precise intra-class localization to pinpoint objects and effective inter-class reasoning to model complex category dependencies. While Vision-Language Pre-training (VLP) models offer a strong open-vocabulary foundation, they often struggle with fine-grained localization under weak supervision and typically fail to explicitly leverage structured relational knowledge beyond basic semantics, limiting performance especially for unseen classes. To overcome these limitations, we propose the Dual Adaptive Refinement Transfer (DART) framework. DART enhances a frozen VLP backbone via two synergistic adaptive modules. For intra-class refinement, an Adaptive Refinement Module (ARM) refines patch features adaptively, coupled with a novel Weakly Supervised Patch Selecting (WPS) loss that enables discriminative localization using only image-level labels. Concurrently, for inter-class transfer, an Adaptive Transfer Module (ATM) leverages a Class Relationship Graph (CRG), constructed using structured knowledge mined from a Large Language Model (LLM), and employs graph attention network to adaptively transfer relational information between class representations. DART is the first framework, to our knowledge, to explicitly integrate external LLM-derived relational knowledge for adaptive inter-class transfer while simultaneously performing adaptive intra-class refinement under weak supervision for OV-MLR. Extensive experiments on challenging benchmarks demonstrate that our DART achieves new state-of-the-art performance, validating its effectiveness.

cs.CV↗

arXiv2022

First Identification of New X-Ray Spectra of Mo39+, Mo40+, W43+, W44+ and W45+ on EAST

New high-resolution x-ray spectra of Mo39+, Mo40+, W43+, W44+ and W45+ have been carefully confirmed for the first time by use of the x-ray imaging crystal spectrometer (XCS) in Experimental Advanced Superconducting Tokamak (EAST) under various combined auxiliary heating plasmas conditions. Wavelength of these new x-ray spectra is ranged from 3.895 Å to 3.986 Å. When core electron temperature (Te0) reaches 6.0 keV, Mo39+ and Mo40+ lines of 3.9727, 3.9294 and 3.9480 Å can be effectively detected on XCS for EAST; meanwhile, line-integrated brightness of these spectral lines of Mo39+ and Mo40+ is very considerable when electron temperature reaches 12.9 keV. Multi-components spectral lines for W43+, W44+ and W45+ have also been identified when Te0 reaches 6 keV. Parts of spectral lines, such as Zn-1, Cu-2, Cu-4a, Cu-4d and Cu-5 lines of tungsten, are first observed experimentally. When electron temperature reaches 12.9 keV, line-integrated intensity for part of these spectral lines of W43+, W44+ and W45+ are considerable. These experimental results and theoretical predictions from FAC and FLYCHK codes are in good general agreement. These new spectral lines, obtained on XCS for EAST, are vital for deeply uncovering the mechanisms of ion and electron thermal, high-Z impurity and momentum (anomalous) transport to achieve the advanced steady-state operation scenarios for ITER and CFETR.

physics.plasm-ph↗

arXiv2017

Using Argument-based Features to Predict and Analyse Review Helpfulness

We study the helpful product reviews identification problem in this paper. We observe that the evidence-conclusion discourse relations, also known as arguments, often appear in product reviews, and we hypothesise that some argument-based features, e.g. the percentage of argumentative sentences, the evidences-conclusions ratios, are good indicators of helpful reviews. To validate this hypothesis, we manually annotate arguments in 110 hotel reviews, and investigate the effectiveness of several combinations of argument-based features. Experiments suggest that, when being used together with the argument-based features, the state-of-the-art baseline features can enjoy a performance boost (in terms of F1) of 11.01\% in average.

cs.CL↗

arXiv2017

Crowdsourcing Argumentation Structures in Chinese Hotel Reviews

Argumentation mining aims at automatically extracting the premises-claim discourse structures in natural language texts. There is a great demand for argumentation corpora for customer reviews. However, due to the controversial nature of the argumentation annotation task, there exist very few large-scale argumentation corpora for customer reviews. In this work, we novelly use the crowdsourcing technique to collect argumentation annotations in Chinese hotel reviews. As the first Chinese argumentation dataset, our corpus includes 4814 argument component annotations and 411 argument relation annotations, and its annotations qualities are comparable to some widely used argumentation corpora in other languages.

cs.CL↗

arXiv2017

Joint RNN Model for Argument Component Boundary Detection

Argument Component Boundary Detection (ACBD) is an important sub-task in argumentation mining; it aims at identifying the word sequences that constitute argument components, and is usually considered as the first sub-task in the argumentation mining pipeline. Existing ACBD methods heavily depend on task-specific knowledge, and require considerable human efforts on feature-engineering. To tackle these problems, in this work, we formulate ACBD as a sequence labeling problem and propose a variety of Recurrent Neural Network (RNN) based methods, which do not use domain specific or handcrafted features beyond the relative position of the sentence in the document. In particular, we propose a novel joint RNN model that can predict whether sentences are argumentative or not, and use the predicted results to more precisely detect the argument component boundaries. We evaluate our techniques on two corpora from two different genres; results suggest that our joint RNN model obtain the state-of-the-art performance on both datasets.

cs.CL↗

arXiv2015

Low-Complexity Downlink User Selection for Massive MIMO Systems

In this paper we propose a pair of low-complexity user selection schemes with zero-forcing precoding for multiuser massive MIMO downlink systems, in which the base station is equipped with a large-scale antenna array. First, we derive approximations of the ergodic sum rates of the systems invoking the conventional random user selection (RUS) and the location-dependant user selection (LUS). Then, the optimal number of simultaneously served user equipments (UEs), $K^*$, is investigated to maximize the sum rate approximations. Upon exploiting $K^*$, we develop two user selection schemes, namely $K^*$-RUS and $K^*$-LUS, where $K^*$ UEs are selected either randomly or based on their locations. Both of the proposed schemes are independent of the instantaneous channel state information of small-scale fading, therefore enjoying the same extremely-low computational complexity as that of the conventional RUS scheme. Moreover, both of our proposed schemes achieve significant sum rate improvement over the conventional RUS. In addition, it is worth noting that like the conventional RUS, the $K^*$-RUS achieves good fairness among UEs.

cs.IT↗

arXiv2014

A Novel Scheme for Downlink Opportunistic Interference Alignment

In this paper we propose a downlink codebook-based opportunistic interference alignment (OIA) in a three-cell MIMO system. A codebook composed of multiple transmit vector sets is utilized to improve the multiuser selection diversity. The sum rate increases as the size of the codebook grows. In addition, during the user selection, effective channel gain and alignment metric are combined to generate a novel criterion, which improves the system performance, especially at low SNR. Furthermore, a threshold-based feedback approach is introduced to reduce the feedback load in the proposed scheme. Both the analytical results and simulations show that the proposed scheme provides significant improvement in terms of sum rates with no feedback load growth and slight increase of complexity.

cs.IT↗

arXiv2014

Low-Complexity Transmission Mode Selection in MU-MIMO Systems

We propose a low-complexity transmission strategy in multi-user multiple-input multiple-output downlink systems. The adaptive strategy adjusts the precoding methods, denoted as the transmission mode, to improve the system sum rates while maintaining the number of simultaneously served users. Three linear precoding transmission modes are discussed, i.e., the block diagonalization zero-forcing, the cooperative zero-forcing (CZF), and the cooperative matched-filter (CMF). Considering both the number of data streams and the multiple-antenna configuration of users, we modify the common CZF and CMF modes by allocating data streams. Then, the transmission mode is selected between the modified ones according to the asymptotic sum rate analyses. As instantaneous channel state information is not needed for the mode selection, the computational complexity is significantly reduced. Numerical simulations confirm our analyses and demonstrate that the proposed scheme achieves substantial performance gains with very low computational complexity.

cs.IT↗