Wen Wen

Fangchen Song, Ashish Agarwal, Wen Wen

Generative artificial intelligence (AI) facilitates content production and enhances ideation capabilities, which can significantly influence developer productivity and participation in software development. To explore its impact on collaborative open-source software (OSS) development, we investigate the role of GitHub Copilot, a generative AI pair programmer, in OSS development where multiple distributed developers voluntarily collaborate. Using GitHub's proprietary Copilot usage data, combined with public OSS project data obtained from GitHub, we find that Copilot use increases project-level code contributions by 5.9%. This gain is driven by a 3.4% rise in developer coding participation and a 2.1% increase in individual productivity. However, Copilot use also leads to an increase in coordination time by 8% due to more code discussions. This reveals an important tradeoff: While AI expands who can contribute and how much they contribute, it slows coordination in collective development efforts. Despite this tension, the combined effect of these two competing forces remains positive, indicating a net gain in overall project-level timely merge of code contributions from using AI pair programmers. Interestingly, we also find the effects differ across developer roles. Peripheral developers show relatively smaller increases in project-level code contributions and experience larger increases in coordination time than core developers. In summary, our study underscores the dual role of AI pair programmers in affecting project-level code contributions and coordination time in OSS development. Our findings on the differential effects between core and peripheral developers also provide important implications for the structure of OSS communities in the long run.

Wen Wen, Mu Li, Yiru Yao et al.

Investigating how people perceive virtual reality (VR) videos in the wild (i.e., those captured by everyday users) is a crucial and challenging task in VR-related applications due to complex authentic distortions localized in space and time. Existing panoramic video databases only consider synthetic distortions, assume fixed viewing conditions, and are limited in size. To overcome these shortcomings, we construct the VR Video Quality in the Wild (VRVQW) database, containing $502$ user-generated videos with diverse content and distortion characteristics. Based on VRVQW, we conduct a formal psychophysical experiment to record the scanpaths and perceived quality scores from $139$ participants under two different viewing conditions. We provide a thorough statistical analysis of the recorded data, observing significant impact of viewing conditions on both human scanpaths and perceived quality. Moreover, we develop an objective quality assessment model for VR videos based on pseudocylindrical representation and convolution. Results on the proposed VRVQW show that our method is superior to existing video quality assessment models. We have made the database and code available at https://github.com/limuhit/VR-Video-Quality-in-the-Wild.

Jing Peng, Junhao Du, Chenghao Wang et al.

Speech foundation models and Speech LLMs have advanced speech understanding, yet deployment-oriented model selection is hindered by non-comparable evaluations caused by mismatched post-processing, and by training results that are hard to reproduce across data scales and pipelines. We present SURE, a unified experimentation framework that standardizes prediction formats, normalization, and scoring. SURE evaluates strong systems across paradigms, from conventional pipelines to Speech LLMs, on representative tasks under realistic acoustic and linguistic stressors. Beyond evaluation, SURE introduces an agent-assisted training conversion flow that maps paper and code into versioned, runnable training pipelines under a unified protocol on matched open-data subsets. Overall, SURE improves comparability and reproducibility for deployment-oriented evaluation.

Emman Haider, Daniel Perez-Becker, Thomas Portet et al. · microsoft-research

Recent innovations in language model training have demonstrated that it is possible to create highly performant models that are small enough to run on a smartphone. As these models are deployed in an increasing number of domains, it is critical to ensure that they are aligned with human preferences and safety considerations. In this report, we present our methodology for safety aligning the Phi-3 series of language models. We utilized a "break-fix" cycle, performing multiple rounds of dataset curation, safety post-training, benchmarking, red teaming, and vulnerability identification to cover a variety of harm areas in both single and multi-turn scenarios. Our results indicate that this approach iteratively improved the performance of the Phi-3 models across a wide range of responsible AI benchmarks. Finally, we include additional red teaming strategies and evaluations that were used to test the safety behavior of Phi-3.5-mini and Phi-3.5-MoE, which were optimized for multilingual capabilities.

Wei Sun, Wen Wen, Xiongkuo Min et al.

Blind video quality assessment (BVQA) plays an indispensable role in monitoring and improving the end-users' viewing experience in various real-world video-enabled media applications. As an experimental field, the improvements of BVQA models have been measured primarily on a few human-rated VQA datasets. Thus, it is crucial to gain a better understanding of existing VQA datasets in order to properly evaluate the current progress in BVQA. Towards this goal, we conduct a first-of-its-kind computational analysis of VQA datasets via designing minimalistic BVQA models. By minimalistic, we restrict our family of BVQA models to build only upon basic blocks: a video preprocessor (for aggressive spatiotemporal downsampling), a spatial quality analyzer, an optional temporal quality analyzer, and a quality regressor, all with the simplest possible instantiations. By comparing the quality prediction performance of different model variants on eight VQA datasets with realistic distortions, we find that nearly all datasets suffer from the easy dataset problem of varying severity, some of which even admit blind image quality assessment (BIQA) solutions. We additionally justify our claims by contrasting our model generalizability on these VQA datasets, and by ablating a dizzying set of BVQA design choices related to the basic building blocks. Our results cast doubt on the current progress in BVQA, and meanwhile shed light on good practices of constructing next-generation VQA datasets and models.

Adarsh Kumarappan, Pareesa Ameneh Golnari, Wen Wen et al.

DevBench is a telemetry-driven benchmark designed to evaluate Large Language Models (LLMs) on realistic code completion tasks. It includes 1,800 evaluation instances across six programming languages and six task categories derived from real developer telemetry and synthesized using generator models from multiple provider families to mitigate single-source bias. Unlike prior benchmarks, it emphasizes ecological validity, avoids training data contamination, and enables detailed diagnostics. The evaluation combines functional correctness, similarity-based metrics, and LLM-judge assessments focused on usefulness and contextual relevance. 9 state-of-the-art models were assessed, with the strongest achieving only 43.5% Pass@1, confirming the benchmark remains challenging and revealing differences in syntactic precision, semantic reasoning, and practical utility. Our benchmark provides actionable insights to guide model selection and improvement, detail that is often missing from other benchmarks but is essential for both practical deployment and targeted model development.

Marah Abdin, Jyoti Aneja, Hany Awadalla et al. · microsoft-research, stanford

We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing, rivals that of models such as Mixtral 8x7B and GPT-3.5 (e.g., phi-3-mini achieves 69% on MMLU and 8.38 on MT-bench), despite being small enough to be deployed on a phone. Our training dataset is a scaled-up version of the one used for phi-2, composed of heavily filtered publicly available web data and synthetic data. The model is also further aligned for robustness, safety, and chat format. We also provide parameter-scaling results with a 7B, 14B models trained for 4.8T tokens, called phi-3-small, phi-3-medium, both significantly more capable than phi-3-mini (e.g., respectively 75%, 78% on MMLU, and 8.7, 8.9 on MT-bench). To enhance multilingual, multimodal, and long-context capabilities, we introduce three models in the phi-3.5 series: phi-3.5-mini, phi-3.5-MoE, and phi-3.5-Vision. The phi-3.5-MoE, a 16 x 3.8B MoE model with 6.6 billion active parameters, achieves superior performance in language reasoning, math, and code tasks compared to other open-source models of similar scale, such as Llama 3.1 and the Mixtral series, and on par with Gemini-1.5-Flash and GPT-4o-mini. Meanwhile, phi-3.5-Vision, a 4.2 billion parameter model derived from phi-3.5-mini, excels in reasoning tasks and is adept at handling both single-image and text prompts, as well as multi-image and text prompts.

Wen Wen, Hao Chen, Shiliang Zhang

Lifelong person re-identification (LReID) aims to train a generalizable model with sequentially collected data. However, such models often suffer from semantic drift, limited adaptability, and catastrophic forgetting as new domains emerge. Existing exemplar-free approaches largely rely on visual-only distillation or parameter regularization, while overlooking the potential of auxiliary modalities, such as text, to preserve semantic stability and enable incremental plasticity. We observe that the frozen text encoder in pretrained vision-language models can serve as a stable semantic anchor across domains. To decouple the roles of vision and text, we propose Prompt-Anchored vision-text Distillation (PAD), an asymmetric vision-text framework for semantic alignment and cross-domain generalization. On the textual side, we distill prompts to preserve vision-text alignment under a fixed semantic space, acting as a global semantic reference rather than a dominant learning signal. On the visual side, an EMA-based teacher with an adaptive prompt pool enables domain-wise adaptation by allocating new slots while freezing past ones. Extensive experiments show that PAD substantially outperforms state-of-the-art methods across seen and unseen domains, achieving a strong balance between stability and plasticity. Project page is available at https://github.com/zu-zi/PAD.

Xin Li, Daoli Xu, Wei Luo et al.

This paper reviews the LoViF 2026 Challenge on Human-oriented Semantic Image Quality Assessment. This challenge aims to raise a new direction, i.e., how to evaluate the loss of semantic information from the human perspective, intending to promote the development of some new directions, like semantic coding, processing, and semantic-oriented optimization, etc. Unlike existing datasets of quality assessment, we form a dataset of human-oriented semantic quality assessment, termed the SeIQA dataset. This dataset is divided into three parts for this competition: (i) training data: 510 pairs of degraded images and their corresponding ground truth references; (ii) validation data: 80 pairs of degraded images and their corresponding ground-truth references; (iii) testing data: 160 pairs of degraded images and their corresponding ground-truth references. The primary objective of this challenge is to establish a new and powerful benchmark for human-oriented semantic image quality assessment. There are a total of 58 teams registered in this competition, and 6 teams submitted valid solutions and fact sheets for the final testing phase. These submissions achieved state-of-the-art (SOTA) performance on the SeIQA dataset.

Wei Sun, Linhan Cao, Yuqin Cao et al.

The rapid proliferation of user-generated content (UGC) on short-form video platforms has made video engagement prediction increasingly important for optimizing recommendation systems and guiding content creation. However, this task remains challenging due to the complex interplay of factors such as semantic content, visual quality, audio characteristics, and user background. Prior studies have leveraged various types of features from different modalities, such as visual quality, semantic content, background sound, etc., but often struggle to effectively model their cross-feature and cross-modality interactions. In this work, we empirically investigate the potential of large multimodal models (LMMs) for video engagement prediction. We adopt two representative LMMs: VideoLLaMA2, which integrates audio, visual, and language modalities, and Qwen2.5-VL, which models only visual and language modalities. Specifically, VideoLLaMA2 jointly processes key video frames, text-based metadata, and background sound, while Qwen2.5-VL utilizes only key video frames and text-based metadata. Trained on the SnapUGC dataset, both models demonstrate competitive performance against state-of-the-art baselines, showcasing the effectiveness of LMMs in engagement prediction. Notably, VideoLLaMA2 consistently outperforms Qwen2.5-VL, highlighting the importance of audio features in engagement prediction. By ensembling two types of models, our method achieves first place in the ICCV VQualA 2025 EVQA-SnapUGC Challenge on short-form video engagement prediction. The code is available at https://github.com/sunwei925/LMM-EVQA.git.

Kanglong Fan, Tianhe Wu, Wen Wen et al.

Reasoning-induced vision-language models (VLMs) advance image quality assessment (IQA) with textual reasoning, yet their scalar scores often lack sensitivity and collapse to a few values, so-called discrete collapse. We introduce ME-IQA, a plug-and-play, test-time memory-enhanced re-ranking framework. It (i) builds a memory bank and retrieves semantically and perceptually aligned neighbors using reasoning summaries, (ii) reframes the VLM as a probabilistic comparator to obtain pairwise preference probabilities and fuse this ordinal evidence with the initial score under Thurstone's Case V model, and (iii) performs gated reflection and consolidates memory to improve future decisions. This yields denser, distortion-sensitive predictions and mitigates discrete collapse. Experiments across multiple IQA benchmarks show consistent gains over strong reasoning-induced VLM baselines, existing non-reasoning IQA methods, and test-time scaling alternatives.

Yanan Chen, Tieliang Gong, Yunjiao Zhang et al.

Continual Learning (CL) aims to enable models to sequentially learn multiple tasks without forgetting previous knowledge. Recent studies have shown that optimizing towards flatter loss minima can improve model generalization. However, existing sharpness-aware methods for CL suffer from two key limitations: (1) they treat sharpness regularization as a unified signal without distinguishing the contributions of its components. and (2) they introduce substantial computational overhead that impedes practical deployment. To address these challenges, we propose FLAD, a novel optimization framework that decomposes sharpness-aware perturbations into gradient-aligned and stochastic-noise components, and show that retaining only the noise component promotes generalization. We further introduce a lightweight scheduling scheme that enables FLAD to maintain significant performance gains even under constrained training time. FLAD can be seamlessly integrated into various CL paradigms and consistently outperforms standard and sharpness-aware optimizers in diverse experimental settings, demonstrating its effectiveness and practicality in CL.

Wen Wen, Mu Li, Yabin Zhang et al.

Blind video quality assessment (BVQA) plays a pivotal role in evaluating and improving the viewing experience of end-users across a wide range of video-based platforms and services. Contemporary deep learning-based models primarily analyze video content in its aggressively subsampled format, while being blind to the impact of the actual spatial resolution and frame rate on video quality. In this paper, we propose a modular BVQA model and a method of training it to improve its modularity. Our model comprises a base quality predictor, a spatial rectifier, and a temporal rectifier, responding to the visual content and distortion, spatial resolution, and frame rate changes on video quality, respectively. During training, spatial and temporal rectifiers are dropped out with some probabilities to render the base quality predictor a standalone BVQA model, which should work better with the rectifiers. Extensive experiments on both professionally-generated content and user-generated content video databases show that our quality model achieves superior or comparable performance to current methods. Additionally, the modularity of our model offers an opportunity to analyze existing video quality databases in terms of their spatial and temporal complexity.

Dasong Li, Sizhuo Ma, Hang Hua et al.

This paper presents an overview of the VQualA 2025 Challenge on Engagement Prediction for Short Videos, held in conjunction with ICCV 2025. The challenge focuses on understanding and modeling the popularity of user-generated content (UGC) short videos on social media platforms. To support this goal, the challenge uses a new short-form UGC dataset featuring engagement metrics derived from real-world user interactions. This objective of the Challenge is to promote robust modeling strategies that capture the complex factors influencing user engagement. Participants explored a variety of multi-modal features, including visual content, audio, and metadata provided by creators. The challenge attracted 97 participants and received 15 valid test submissions, contributing significantly to progress in short-form UGC video engagement prediction.

Kanglong Fan, Wen Wen, Mu Li et al.

Panoramic videos have the advantage of providing an immersive and interactive viewing experience. Nevertheless, their spherical nature gives rise to various and uncertain user viewing behaviors, which poses significant challenges for panoramic video quality assessment (PVQA). In this work, we propose an end-to-end optimized, blind PVQA method with explicit modeling of user viewing patterns through visual scanpaths. Our method consists of two modules: a scanpath generator and a quality assessor. The scanpath generator is initially trained to predict future scanpaths by minimizing their expected code length and then jointly optimized with the quality assessor for quality prediction. Our blind PVQA method enables direct quality assessment of panoramic images by treating them as videos composed of identical frames. Experiments on three public panoramic image and video quality datasets, encompassing both synthetic and authentic distortions, validate the superiority of our blind PVQA model over existing methods.

Wen Wen, Yilin Wang, Neil Birkbeck et al.

The rise of short-form videos, characterized by diverse content, editing styles, and artifacts, poses substantial challenges for learning-based blind video quality assessment (BVQA) models. Multimodal large language models (MLLMs), renowned for their superior generalization capabilities, present a promising solution. This paper focuses on effectively leveraging a pretrained MLLM for short-form video quality assessment, regarding the impacts of pre-processing and response variability, and insights on combining the MLLM with BVQA models. We first investigated how frame pre-processing and sampling techniques influence the MLLM's performance. Then, we introduced a lightweight learning-based ensemble method that adaptively integrates predictions from the MLLM and state-of-the-art BVQA models. Our results demonstrated superior generalization performance with the proposed ensemble approach. Furthermore, the analysis of content-aware ensemble weights highlighted that some video characteristics are not fully represented by existing BVQA models, revealing potential directions to improve BVQA models further.

Weifan Long, Wen Wen, Peng Zhai et al.

Zero-shot coordination problem in multi-agent reinforcement learning (MARL), which requires agents to adapt to unseen agents, has attracted increasing attention. Traditional approaches often rely on the Self-Play (SP) framework to generate a diverse set of policies in a policy pool, which serves to improve the generalization capability of the final agent. However, these frameworks may struggle to capture the full spectrum of potential strategies, especially in real-world scenarios that demand agents balance cooperation with competition. In such settings, agents need strategies that can adapt to varying and often conflicting goals. Drawing inspiration from Social Value Orientation (SVO)-where individuals maintain stable value orientations during interactions with others-we propose a novel framework called \emph{Role Play} (RP). RP employs role embeddings to transform the challenge of policy diversity into a more manageable diversity of roles. It trains a common policy with role embedding observations and employs a role predictor to estimate the joint role embeddings of other agents, helping the learning agent adapt to its assigned role. We theoretically prove that an approximate optimal policy can be achieved by optimizing the expected cumulative reward relative to an approximate role-based policy. Experimental results in both cooperative (Overcooked) and mixed-motive games (Harvest, CleanUp) reveal that RP consistently outperforms strong baselines when interacting with unseen agents, highlighting its robustness and adaptability in complex environments.

Wen Wen, Tieliang Gong, Yunjiao Zhang et al.

Continual learning (CL) has emerged as a dominant paradigm for acquiring knowledge from sequential tasks while avoiding catastrophic forgetting. Although many CL methods have been proposed to show impressive empirical performance, the theoretical understanding of their generalization behavior remains limited, particularly for replay-based approaches. In this paper, we establish a unified theoretical framework for replay-based CL, deriving a series of information-theoretic bounds that explicitly characterize how the memory buffer interacts with the current task to affect generalization. Specifically, our hypothesis-based bounds reveal that utilizing the limited exemplars of previous tasks alongside the current task data, rather than exhaustive replay, facilitates improved generalization while effectively mitigating catastrophic forgetting. Furthermore, our prediction-based bounds yield tighter and computationally tractable upper bounds of the generalization gap through the use of low-dimensional variables. Our analysis is general and broadly applicable to a wide range of learning algorithms, exemplified by stochastic gradient Langevin dynamics (SGLD) as a representative method. Comprehensive experimental evaluations demonstrate the effectiveness of our derived bounds in capturing the generalization dynamics in replay-based CL settings.

Wen Wen, Yaohong Wu, Yue Sheng et al.

Video quality assessment (VQA) is a challenging research topic with broad applications. Effective VQA necessitates sensitivity to pixel-level distortions and a comprehensive understanding of video context to accurately determine the perceptual impact of distortions. Traditional hand-crafted and learning-based VQA models mainly focus on pixel-level distortions and lack contextual understanding, while recent LLM-based models struggle with sensitivity to small distortions or handle quality scoring and description as separate tasks. To address these shortcomings, we introduce CP-LLM: a Context and Pixel aware Large Language Model. CP-LLM is a novel multimodal LLM architecture featuring dual vision encoders designed to independently analyze perceptual quality at both high-level (video context) and low-level (pixel distortion) granularity, along with a language decoder subsequently reasons about the interplay between these aspects. This design enables CP-LLM to simultaneously produce robust quality scores and interpretable quality descriptions, with enhanced sensitivity to pixel distortions (e.g. compression artifacts). The model is trained via a multi-task pipeline optimizing for score prediction, description generation, and pairwise comparisons. Experiment results demonstrate that CP-LLM achieves state-of-the-art cross-dataset performance on established VQA benchmarks and superior robustness to pixel distortions, confirming its efficacy for comprehensive and practical video quality assessment in real-world scenarios.

Wen Wen, Tianwu Zhi, Kanglong Fan et al.

Improving vision-language models (VLMs) in the post-training stage typically relies on supervised fine-tuning or reinforcement learning, methods that necessitate costly, human-annotated data. While self-supervised techniques such as self-consistency have proven effective for enhancing reasoning capabilities, their application to perceptual domains such as image quality assessment (IQA) remains largely unexplored. In this work, we introduce EvoQuality, a novel framework that enables a VLM to autonomously refine its quality perception capabilities without any ground-truth labels. EvoQuality adapts the principle of self-consistency to the ranking-based nature of IQA. It generates pseudo-labels by performing pairwise majority voting on the VLM's own outputs to establish a consensus on relative quality. These pseudo-rankings are then formulated into a fidelity reward that guides the model's iterative evolution through group relative policy optimization (GRPO). By iteratively leveraging its own predictions, EvoQuality progressively refines the VLM's perceptual capability. Extensive experiments show that EvoQuality boosts the base VLM's zero-shot performance by 31.8\% on PLCC across diverse IQA benchmarks. Remarkably, despite being entirely self-supervised, EvoQuality achieves performance that is competitive with, or even surpasses, state-of-the-art supervised VLM-based IQA models, outperforming these models on 5 out of 7 IQA benchmarks.

Yuanhong Zhang, Muyao Yuan, Weizhan Zhang et al.

The Segment Anything Model (SAM), a vision foundation model, exhibits impressive zero-shot capabilities in general tasks but struggles in specialized domains. Parameter-efficient fine-tuning (PEFT) is a promising approach to unleash the potential of SAM in novel scenarios. However, existing PEFT methods for SAM neglect the domain-invariant relations encoded in the pre-trained model. To bridge this gap, we propose InfoSAM, an information-theoretic approach that enhances SAM fine-tuning by distilling and preserving its pre-trained segmentation knowledge. Specifically, we formulate the knowledge transfer process as two novel mutual information-based objectives: (i) to compress the domain-invariant relation extracted from pre-trained SAM, excluding pseudo-invariant information as possible, and (ii) to maximize mutual information between the relational knowledge learned by the teacher (pre-trained SAM) and the student (fine-tuned model). The proposed InfoSAM establishes a robust distillation framework for PEFT of SAM. Extensive experiments across diverse benchmarks validate InfoSAM's effectiveness in improving SAM family's performance on real-world tasks, demonstrating its adaptability and superiority in handling specialized scenarios.

Wen Wen, Tieliang Gong, Yuxin Dong et al.

Multiview learning has drawn widespread attention for its efficacy in leveraging cross-view consensus and complementarity information to achieve a comprehensive representation of data. While multi-view learning has undergone vigorous development and achieved remarkable success, the theoretical understanding of its generalization behavior remains elusive. This paper aims to bridge this gap by developing information-theoretic generalization bounds for multi-view learning, with a particular focus on multi-view reconstruction and classification tasks. Our bounds underscore the importance of capturing both consensus and complementary information from multiple different views to achieve maximally disentangled representations. These results also indicate that applying the multi-view information bottleneck regularizer is beneficial for satisfactory generalization performance. Additionally, we derive novel data-dependent bounds under both leave-one-out and supersample settings, yielding computational tractable and tighter bounds. In the interpolating regime, we further establish the fast-rate bound for multi-view learning, exhibiting a faster convergence rate compared to conventional square-root bounds. Numerical results indicate a strong correlation between the true generalization gap and the derived bounds across various learning scenarios.

Wen Wen, Tieliang Gong, Yuxin Dong et al.

In recent years, information-theoretic generalization bounds have gained increasing attention for analyzing the generalization capabilities of meta-learning algorithms. However, existing results are confined to two-step bounds, failing to provide a sharper characterization of the meta-generalization gap that simultaneously accounts for environment-level and task-level dependencies. This paper addresses this fundamental limitation by developing a unified information-theoretic framework using a single-step derivation. The resulting meta-generalization bounds, expressed in terms of diverse information measures, exhibit substantial advantages over previous work, particularly in terms of tightness, scaling behavior associated with sampled tasks and samples per task, and computational tractability. Furthermore, through gradient covariance analysis, we provide new theoretical insights into the generalization properties of two classes of noisy and iterative meta-learning algorithms, where the meta-learner uses either the entire meta-training data (e.g., Reptile), or separate training and test data within the task (e.g., model agnostic meta-learning (MAML)). Numerical results validate the effectiveness of the derived bounds in capturing the generalization dynamics of meta-learning.

Qing Lin, Yongbin Liu, Wen Wen et al.

Few-shot learning has been proposed and rapidly emerging as a viable means for completing various tasks. Many few-shot models have been widely used for relation learning tasks. However, each of these models has a shortage of capturing a certain aspect of semantic features, for example, CNN on long-range dependencies part, Transformer on local features. It is difficult for a single model to adapt to various relation learning, which results in the high variance problem. Ensemble strategy could be competitive on improving the accuracy of few-shot relation extraction and mitigating high variance risks. This paper explores an ensemble approach to reduce the variance and introduces fine-tuning and feature attention strategies to calibrate relation-level features. Results on several few-shot relation learning tasks show that our model significantly outperforms the previous state-of-the-art models.

Zhifeng Hao, Di Lv, Zijian Li et al.

Named entity recognition (NER) for identifying proper nouns in unstructured text is one of the most important and fundamental tasks in natural language processing. However, despite the widespread use of NER models, they still require a large-scale labeled data set, which incurs a heavy burden due to manual annotation. Domain adaptation is one of the most promising solutions to this problem, where rich labeled data from the relevant source domain are utilized to strengthen the generalizability of a model based on the target domain. However, the mainstream cross-domain NER models are still affected by the following two challenges (1) Extracting domain-invariant information such as syntactic information for cross-domain transfer. (2) Integrating domain-specific information such as semantic information into the model to improve the performance of NER. In this study, we present a semi-supervised framework for transferable NER, which disentangles the domain-invariant latent variables and domain-specific latent variables. In the proposed framework, the domain-specific information is integrated with the domain-specific latent variables by using a domain predictor. The domain-specific and domain-invariant latent variables are disentangled using three mutual information regularization terms, i.e., maximizing the mutual information between the domain-specific latent variables and the original embedding, maximizing the mutual information between the domain-invariant latent variables and the original embedding, and minimizing the mutual information between the domain-specific and domain-invariant latent variables. Extensive experiments demonstrated that our model can obtain state-of-the-art performance with cross-domain and cross-lingual NER benchmark data sets.