Li Chen

Yihan Hu, Jiazhi Yang, Li Chen et al.

Modern autonomous driving system is characterized as modular tasks in sequential order, i.e., perception, prediction, and planning. In order to perform a wide diversity of tasks and achieve advanced-level intelligence, contemporary approaches either deploy standalone models for individual tasks, or design a multi-task paradigm with separate heads. However, they might suffer from accumulative errors or deficient task coordination. Instead, we argue that a favorable framework should be devised and optimized in pursuit of the ultimate goal, i.e., planning of the self-driving car. Oriented at this, we revisit the key components within perception and prediction, and prioritize the tasks such that all these tasks contribute to planning. We introduce Unified Autonomous Driving (UniAD), a comprehensive framework up-to-date that incorporates full-stack driving tasks in one network. It is exquisitely devised to leverage advantages of each module, and provide complementary feature abstractions for agent interaction from a global perspective. Tasks are communicated with unified query interfaces to facilitate each other toward planning. We instantiate UniAD on the challenging nuScenes benchmark. With extensive ablations, the effectiveness of using such a philosophy is proven by substantially outperforming previous state-of-the-arts in all aspects. Code and models are public.

Xiaosong Jia, Yulu Gao, Li Chen et al.

End-to-end autonomous driving aims to build a fully differentiable system that takes raw sensor data as inputs and directly outputs the planned trajectory or control signals of the ego vehicle. State-of-the-art methods usually follow the `Teacher-Student' paradigm. The Teacher model uses privileged information (ground-truth states of surrounding agents and map elements) to learn the driving strategy. The student model only has access to raw sensor data and conducts behavior cloning on the data collected by the teacher model. By eliminating the noise of the perception part during planning learning, state-of-the-art works could achieve better performance with significantly less data compared to those coupled ones. However, under the current Teacher-Student paradigm, the student model still needs to learn a planning head from scratch, which could be challenging due to the redundant and noisy nature of raw sensor inputs and the casual confusion issue of behavior cloning. In this work, we aim to explore the possibility of directly adopting the strong teacher model to conduct planning while letting the student model focus more on the perception part. We find that even equipped with a SOTA perception model, directly letting the student model learn the required inputs of the teacher model leads to poor driving performance, which comes from the large distribution gap between predicted privileged inputs and the ground-truth. To this end, we propose DriveAdapter, which employs adapters with the feature alignment objective function between the student (perception) and teacher (planning) modules. Additionally, since the pure learning-based teacher model itself is imperfect and occasionally breaks safety rules, we propose a method of action-guided feature learning with a mask for those imperfect teacher features to further inject the priors of hand-crafted rules into the learning process.

Yuxin Zhu, Xilei Zhu, Huiyu Duan et al.

Visual saliency prediction for omnidirectional videos (ODVs) has shown great significance and necessity for omnidirectional videos to help ODV coding, ODV transmission, ODV rendering, etc.. However, most studies only consider visual information for ODV saliency prediction while audio is rarely considered despite its significant influence on the viewing behavior of ODV. This is mainly due to the lack of large-scale audio-visual ODV datasets and corresponding analysis. Thus, in this paper, we first establish the largest audio-visual saliency dataset for omnidirectional videos (AVS-ODV), which comprises the omnidirectional videos, audios, and corresponding captured eye-tracking data for three video sound modalities including mute, mono, and ambisonics. Then we analyze the visual attention behavior of the observers under various omnidirectional audio modalities and visual scenes based on the AVS-ODV dataset. Furthermore, we compare the performance of several state-of-the-art saliency prediction models on the AVS-ODV dataset and construct a new benchmark. Our AVS-ODV datasets and the benchmark will be released to facilitate future research.

Yukai Miao, Yu Bai, Li Chen et al.

Nowadays, the versatile capabilities of Pre-trained Large Language Models (LLMs) have attracted much attention from the industry. However, some vertical domains are more interested in the in-domain capabilities of LLMs. For the Networks domain, we present NetEval, an evaluation set for measuring the comprehensive capabilities of LLMs in Network Operations (NetOps). NetEval is designed for evaluating the commonsense knowledge and inference ability in NetOps in a multi-lingual context. NetEval consists of 5,732 questions about NetOps, covering five different sub-domains of NetOps. With NetEval, we systematically evaluate the NetOps capability of 26 publicly available LLMs. The results show that only GPT-4 can achieve a performance competitive to humans. However, some open models like LLaMA 2 demonstrate significant potential.

Yang Liu, Jiaxiang Liu, Li Chen et al.

Sparse Transformer has recently attracted a lot of attention since the ability for reducing the quadratic dependency on the sequence length. We argue that two factors, information bottleneck sensitivity and inconsistency between different attention topologies, could affect the performance of the Sparse Transformer. This paper proposes a well-designed model named ERNIE-Sparse. It consists of two distinctive parts: (i) Hierarchical Sparse Transformer (HST) to sequentially unify local and global information. (ii) Self-Attention Regularization (SAR) method, a novel regularization designed to minimize the distance for transformers with different attention topologies. To evaluate the effectiveness of ERNIE-Sparse, we perform extensive evaluations. Firstly, we perform experiments on a multi-modal long sequence modeling task benchmark, Long Range Arena (LRA). Experimental results demonstrate that ERNIE-Sparse significantly outperforms a variety of strong baseline methods including the dense attention and other efficient sparse attention methods and achieves improvements by 2.77% (57.78% vs. 55.01%). Secondly, to further show the effectiveness of our method, we pretrain ERNIE-Sparse and verified it on 3 text classification and 2 QA downstream tasks, achieve improvements on classification benchmark by 0.83% (92.46% vs. 91.63%), on QA benchmark by 3.24% (74.67% vs. 71.43%). Experimental results continue to demonstrate its superior performance.

Weiyan Wang, Yilun Jin, Yiming Zhang et al.

Due to high accuracy, BERT-like models have been widely adopted by text mining and web searching. However, large BERT-like models suffer from inefficient online inference, facing the following two problems on GPUs: (1) their high accuracy relies on the large model depth, which linearly increases the sequential computation on GPUs; (2) stochastic and dynamic online workloads cause extra costs from batching and paddings. Therefore, we present \sys for the real-world setting of GPU inference on online workloads. At its core, \sys adopts stacking distillation and boosting ensemble, distilling the original deep model into a group of shallow but virtually stacked student models running in parallel. This enables \sys to achieve a lower model depth (e.g., two layers) than the others and the lowest inference latency while maintaining accuracy. In addition, adaptive student pruning realizes dynamic student numbers according to changing online workloads. Especially for occasional workload bursts, it can temporarily decrease the student number with minimal accuracy loss to improve system throughput. We conduct comprehensive experiments to verify the effectiveness, whose results show that \sys outperforms the baselines by $4.1\times\sim 1.6\times$ in latency while maintaining accuracy and achieves up to $22.27\times$ higher throughput for workload bursts.

Jia Zeng, Qingwen Bu, Bangjun Wang et al.

Representation learning approaches for robotic manipulation have boomed in recent years. Due to the scarcity of in-domain robot data, prevailing methodologies tend to leverage large-scale human video datasets to extract generalizable features for visuomotor policy learning. Despite the progress achieved, prior endeavors disregard the interactive dynamics that capture behavior patterns and physical interaction during the manipulation process, resulting in an inadequate understanding of the relationship between objects and the environment. To this end, we propose a general pre-training pipeline that learns Manipulation by Predicting the Interaction (MPI) and enhances the visual representation.Given a pair of keyframes representing the initial and final states, along with language instructions, our algorithm predicts the transition frame and detects the interaction object, respectively. These two learning objectives achieve superior comprehension towards "how-to-interact" and "where-to-interact". We conduct a comprehensive evaluation of several challenging robotic tasks.The experimental results demonstrate that MPI exhibits remarkable improvement by 10% to 64% compared with previous state-of-the-art in real-world robot platforms as well as simulation environments. Code and checkpoints are publicly shared at https://github.com/OpenDriveLab/MPI.

Yizhe Zhang, Yucheng Jin, Li Chen et al.

Conversational recommender systems (CRS) enable users to articulate their preferences and provide feedback through natural language. With the advent of large language models (LLMs), the potential to enhance user engagement with CRS and augment the recommendation process with LLM-generated content has received increasing attention. However, the efficacy of LLM-powered CRS is contingent upon the use of prompts, and the subjective perception of recommendation quality can differ across various recommendation domains. Therefore, we have developed a ChatGPT-based CRS to investigate the impact of these two factors, prompt guidance (PG) and recommendation domain (RD), on the overall user experience of the system. We conducted an online empirical study (N = 100) by employing a mixed-method approach that utilized a between-subjects design for the variable of PG (with vs. without) and a within-subjects design for RD (book recommendations vs. job recommendations). The findings reveal that PG can substantially enhance the system's explainability, adaptability, perceived ease of use, and transparency. Moreover, users are inclined to perceive a greater sense of novelty and demonstrate a higher propensity to engage with and try recommended items in the context of book recommendations as opposed to job recommendations. Furthermore, the influence of PG on certain user experience metrics and interactive behaviors appears to be modulated by the recommendation domain, as evidenced by the interaction effects between the two examined factors. This work contributes to the user-centered evaluation of ChatGPT-based CRS by investigating two prominent factors and offers practical design guidance.

Qichen Sun, Yuxing Lu, Kun Xia et al. · pku

Rapid and efficient assessment of the future impact of research articles is a significant concern for both authors and reviewers. The most common standard for measuring the impact of academic papers is the number of citations. In recent years, numerous efforts have been undertaken to predict citation counts within various citation windows. However, most of these studies focus solely on a specific academic field or require early citation counts for prediction, rendering them impractical for the early-stage evaluation of papers. In this work, we harness Scopus to curate a significantly comprehensive and large-scale dataset of information from 69707 scientific articles sourced from 99 journals spanning multiple disciplines. We propose a deep learning methodology for the impact-based classification tasks, which leverages semantic features extracted from the manuscripts and paper metadata. To summarize the semantic features, such as titles and abstracts, we employ a Transformer-based language model to encode semantic features and design a text fusion layer to capture shared information between titles and abstracts. We specifically focus on the following impact-based prediction tasks using information of scientific manuscripts in pre-publication stage: (1) The impact of journals in which the manuscripts will be published. (2) The future impact of manuscripts themselves. Extensive experiments on our datasets demonstrate the superiority of our proposed model for impact-based prediction tasks. We also demonstrate potentials in generating manuscript's feedback and improvement suggestions.

Lei Li, Yongfeng Zhang, Dugang Liu et al.

Large language models (LLM) not only have revolutionized the field of natural language processing (NLP) but also have the potential to reshape many other fields, e.g., recommender systems (RS). However, most of the related work treats an LLM as a component of the conventional recommendation pipeline (e.g., as a feature extractor), which may not be able to fully leverage the generative power of LLM. Instead of separating the recommendation process into multiple stages, such as score computation and re-ranking, this process can be simplified to one stage with LLM: directly generating recommendations from the complete pool of items. This survey reviews the progress, methods, and future directions of LLM-based generative recommendation by examining three questions: 1) What generative recommendation is, 2) Why RS should advance to generative recommendation, and 3) How to implement LLM-based generative recommendation for various RS tasks. We hope that this survey can provide the context and guidance needed to explore this interesting and emerging topic.

Lei Li, Yongfeng Zhang, Li Chen

Providing user-understandable explanations to justify recommendations could help users better understand the recommended items, increase the system's ease of use, and gain users' trust. A typical approach to realize it is natural language generation. However, previous works mostly adopt recurrent neural networks to meet the ends, leaving the potentially more effective pre-trained Transformer models under-explored. In fact, user and item IDs, as important identifiers in recommender systems, are inherently in different semantic space as words that pre-trained models were already trained on. Thus, how to effectively fuse IDs into such models becomes a critical issue. Inspired by recent advancement in prompt learning, we come up with two solutions: find alternative words to represent IDs (called discrete prompt learning), and directly input ID vectors to a pre-trained model (termed continuous prompt learning). In the latter case, ID vectors are randomly initialized but the model is trained in advance on large corpora, so they are actually in different learning stages. To bridge the gap, we further propose two training strategies: sequential tuning and recommendation as regularization. Extensive experiments show that our continuous prompt learning approach equipped with the training strategies consistently outperforms strong baselines on three datasets of explainable recommendation.

Dongyang Dai, Yuanzhe Chen, Li Chen et al.

With the increasing popularity of speech synthesis products, the industry has put forward more requirements for personalized speech synthesis: (1) How to use low-resource, easily accessible data to clone a person's voice. (2) How to clone a person's voice while controlling the style and prosody. To solve the above two problems, we proposed the Hieratron model framework in which the prosody and timbre are modeled separately using two modules, therefore, the independent control of timbre and the other characteristics of audio can be achieved while generating speech. The practice shows that, for very limited target speaker data in the wild, Hieratron has obvious advantages over the traditional method, in addition to controlling the style and language of the generated speech, the mean opinion score on speech quality of the generated speech has also been improved by more than 0.2 points.

Lingmei Bu, Li Chen, Yongmei Lu et al.

Extracting aspect-polarity pairs from texts is an important task of fine-grained sentiment analysis. While the existing approaches to this task have gained many progresses, they are limited at capturing relationships among aspect-polarity pairs in a text, thus degrading the extraction performance. Moreover, the existing state-of-the-art approaches, namely token-based se-quence tagging and span-based classification, have their own defects such as polarity inconsistency resulted from separately tagging tokens in the former and the heterogeneous categorization in the latter where aspect-related and polarity-related labels are mixed. In order to remedy the above defects, in-spiring from the recent advancements in relation extraction, we propose to generate aspect-polarity pairs directly from a text with relation extraction technology, regarding aspect-pairs as unary relations where aspects are enti-ties and the corresponding polarities are relations. Based on the perspective, we present a position- and aspect-aware sequence2sequence model for joint extraction of aspect-polarity pairs. The model is characterized with its ability to capture not only relationships among aspect-polarity pairs in a text through the sequence decoding, but also correlations between an aspect and its polarity through the position- and aspect-aware attentions. The experi-ments performed on three benchmark datasets demonstrate that our model outperforms the existing state-of-the-art approaches, making significant im-provement over them.

Lei Li, Yongfeng Zhang, Li Chen

Personalization of natural language generation plays a vital role in a large spectrum of tasks, such as explainable recommendation, review summarization and dialog systems. In these tasks, user and item IDs are important identifiers for personalization. Transformer, which is demonstrated with strong language modeling capability, however, is not personalized and fails to make use of the user and item IDs since the ID tokens are not even in the same semantic space as the words. To address this problem, we present a PErsonalized Transformer for Explainable Recommendation (PETER), on which we design a simple and effective learning objective that utilizes the IDs to predict the words in the target explanation, so as to endow the IDs with linguistic meanings and to achieve personalized Transformer. Besides generating explanations, PETER can also make recommendations, which makes it a unified model for the whole recommendation-explanation pipeline. Extensive experiments show that our small unpretrained model outperforms fine-tuned BERT on the generation task, in terms of both effectiveness and efficiency, which highlights the importance and the nice utility of our design.

Lei Li, Yongfeng Zhang, Li Chen

Recently, research on explainable recommender systems has drawn much attention from both academia and industry, resulting in a variety of explainable models. As a consequence, their evaluation approaches vary from model to model, which makes it quite difficult to compare the explainability of different models. To achieve a standard way of evaluating recommendation explanations, we provide three benchmark datasets for EXplanaTion RAnking (denoted as EXTRA), on which explainability can be measured by ranking-oriented metrics. Constructing such datasets, however, poses great challenges. First, user-item-explanation triplet interactions are rare in existing recommender systems, so how to find alternatives becomes a challenge. Our solution is to identify nearly identical sentences from user reviews. This idea then leads to the second challenge, i.e., how to efficiently categorize the sentences in a dataset into different groups, since it has quadratic runtime complexity to estimate the similarity between any two sentences. To mitigate this issue, we provide a more efficient method based on Locality Sensitive Hashing (LSH) that can detect near-duplicates in sub-linear time for a given query. Moreover, we make our code publicly available to allow researchers in the community to create their own datasets.

Lei Li, Yongfeng Zhang, Li Chen

Explaining to users why some items are recommended is critical, as it can help users to make better decisions, increase their satisfaction, and gain their trust in recommender systems (RS). However, existing explainable RS usually consider explanation as a side output of the recommendation model, which has two problems: (1) it is difficult to evaluate the produced explanations because they are usually model-dependent, and (2) as a result, how the explanations impact the recommendation performance is less investigated. In this paper, explaining recommendations is formulated as a ranking task, and learned from data, similar to item ranking for recommendation. This makes it possible for standard evaluation of explanations via ranking metrics (e.g., NDCG). Furthermore, this paper extends traditional item ranking to an item-explanation joint-ranking formalization to study if purposely selecting explanations could reach certain learning goals, e.g., improving recommendation performance. A great challenge, however, is that the sparsity issue in the user-item-explanation data would be inevitably severer than that in traditional user-item interaction data, since not every user-item pair can be associated with all explanations. To mitigate this issue, this paper proposes to perform two sets of matrix factorization by considering the ternary relationship as two groups of binary relationships. Experiments on three large datasets verify the solution's effectiveness on both explanation ranking and item recommendation.

Zhe Peng, Jinbin Huang, Haixin Wang et al.

The coronavirus disease 2019 (COVID-19) pandemic has caused an unprecedented health crisis for the global. Digital contact tracing, as a transmission intervention measure, has shown its effectiveness on pandemic control. Despite intensive research on digital contact tracing, existing solutions can hardly meet users' requirements on privacy and convenience. In this paper, we propose BU-Trace, a novel permissionless mobile system for privacy-preserving intelligent contact tracing based on QR code and NFC technologies. First, a user study is conducted to investigate and quantify the user acceptance of a mobile contact tracing system. Second, a decentralized system is proposed to enable contact tracing while protecting user privacy. Third, an intelligent behavior detection algorithm is designed to ease the use of our system. We implement BU-Trace and conduct extensive experiments in several real-world scenarios. The experimental results show that BU-Trace achieves a privacy-preserving and intelligent mobile system for contact tracing without requesting location or other privacy-related permissions.

Runxin Xu, Jun Cao, Mingxuan Wang et al.

This paper proposes the building of Xiaomingbot, an intelligent, multilingual and multimodal software robot equipped with four integral capabilities: news generation, news translation, news reading and avatar animation. Its system summarizes Chinese news that it automatically generates from data tables. Next, it translates the summary or the full article into multiple languages, and reads the multilingual rendition through synthesized speech. Notably, Xiaomingbot utilizes a voice cloning technology to synthesize the speech trained from a real person's voice data in one input language. The proposed system enjoys several merits: it has an animated avatar, and is able to generate and read multilingual news. Since it was put into practice, Xiaomingbot has written over 600,000 articles, and gained over 150,000 followers on social media platforms.

Zhi Liu, Yan Huang, Jing Gao et al.

Similar product recommendation is one of the most common scenes in e-commerce. Many recommendation algorithms such as item-to-item Collaborative Filtering are working on measuring item similarities. In this paper, we introduce our real-time personalized algorithm to model product similarity and real-time user interests. We also introduce several other baseline algorithms including an image-similarity-based method, item-to-item collaborative filtering, and item2vec, and compare them on our large-scale real-world e-commerce dataset. The algorithms which achieve good offline results are also tested on the online e-commerce website. Our personalized method achieves a 10% improvement on the add-cart number in the real-world e-commerce scenario.

Li Chen, Ningxia Wang, Yonghua Yang et al.

Though it has been recognized that recommending serendipitous (i.e., surprising and relevant) items can be helpful for increasing users' satisfaction and behavioral intention, how to measure serendipity in the offline environment is still an open issue. In recent years, a number of metrics have been proposed, but most of them were based on researchers' assumptions due to the serendipity's subjective nature. In order to validate these metrics' actual performance, we collected over 10,000 users' real feedback data and compared with the metrics' results. It turns out the user profile based metrics, especially content-based ones, perform better than those based on item popularity, in terms of estimating the unexpectedness facet of recommendations. Moreover, the full metrics, which involve the unexpectedness component, relevance, timeliness, and user curiosity, can more accurately indicate the recommendation's serendipity degree, relative to those that just involve some of them. The application of these metrics to several recommender algorithms further consolidates their practical usage, because the comparison results are consistent with those from user evaluation. Thus, this work is constructive for filling the gap between offline measurement and user study on recommendation serendipity.

Nilaksh Das, Madhuri Shanbhogue, Shang-Tse Chen et al.

Adversarial machine learning research has recently demonstrated the feasibility to confuse automatic speech recognition (ASR) models by introducing acoustically imperceptible perturbations to audio samples. To help researchers and practitioners gain better understanding of the impact of such attacks, and to provide them with tools to help them more easily evaluate and craft strong defenses for their models, we present ADAGIO, the first tool designed to allow interactive experimentation with adversarial attacks and defenses on an ASR model in real time, both visually and aurally. ADAGIO incorporates AMR and MP3 audio compression techniques as defenses, which users can interactively apply to attacked audio samples. We show that these techniques, which are based on psychoacoustic principles, effectively eliminate targeted attacks, reducing the attack success rate from 92.5% to 0%. We will demonstrate ADAGIO and invite the audience to try it on the Mozilla Common Voice dataset.

Wei Zeng, Li Chen

Due to the development of social media technology, it becomes easier for users to gather together to form groups. Take the Last.fm for example, users can join groups they may be interested where they can share their loved songs and discuss topics about songs and singers. However, the number of groups grows over time, users need effective groups recommendations in order to meet more like-minded users.