Chao Bian

Xinnian Liang, Zefan Zhou, Hui Huang et al.

Pretrained language models (PLMs) have shown marvelous improvements across various NLP tasks. Most Chinese PLMs simply treat an input text as a sequence of characters, and completely ignore word information. Although Whole Word Masking can alleviate this, the semantics in words is still not well represented. In this paper, we revisit the segmentation granularity of Chinese PLMs. We propose a mixed-granularity Chinese BERT (MigBERT) by considering both characters and words. To achieve this, we design objective functions for learning both character and word-level representations. We conduct extensive experiments on various Chinese NLP tasks to evaluate existing PLMs as well as the proposed MigBERT. Experimental results show that MigBERT achieves new SOTA performance on all these tasks. Further analysis demonstrates that words are semantically richer than characters. More interestingly, we show that MigBERT also works with Japanese. Our code and model have been released here~\footnote{https://github.com/xnliang98/MigBERT}.

Xinnian Liang, Shuangzhi Wu, Chenhao Cui et al.

Dialogue summarization aims to condense a given dialogue into a simple and focused summary text. Typically, both the roles' viewpoints and conversational topics change in the dialogue stream. Thus how to effectively handle the shifting topics and select the most salient utterance becomes one of the major challenges of this task. In this paper, we propose a novel topic-aware Global-Local Centrality (GLC) model to help select the salient context from all sub-topics. The centralities are constructed at both the global and local levels. The global one aims to identify vital sub-topics in the dialogue and the local one aims to select the most important context in each sub-topic. Specifically, the GLC collects sub-topic based on the utterance representations. And each utterance is aligned with one sub-topic. Based on the sub-topics, the GLC calculates global- and local-level centralities. Finally, we combine the two to guide the model to capture both salient context and sub-topics when generating summaries. Experimental results show that our model outperforms strong baselines on three public dialogue summarization datasets: CSDS, MC, and SAMSUM. Further analysis demonstrates that our GLC can exactly identify vital contents from sub-topics.~\footnote{\url{https://github.com/xnliang98/bart-glc}}

Chao Bian, Yawen Zhou, Miqing Li et al.

Evolutionary algorithms (EAs) have been widely and successfully applied to solve multi-objective optimization problems, due to their nature of population-based search. Population update, a key component in multi-objective EAs (MOEAs), is usually performed in a greedy, deterministic manner. That is, the next-generation population is formed by selecting the best solutions from the current population and newly-generated solutions (irrespective of the selection criteria used such as Pareto dominance, crowdedness and indicators). In this paper, we analytically present that stochastic population update can be beneficial for the search of MOEAs. Specifically, we prove that the expected running time of two well-established MOEAs, SMS-EMOA and NSGA-II, for solving two bi-objective problems, OneJumpZeroJump and bi-objective RealRoyalRoad, can be exponentially decreased if replacing its deterministic population update mechanism by a stochastic one. Empirical studies also verify the effectiveness of the proposed population update method. This work is an attempt to show the benefit of introducing randomness into the population update of MOEAs. Its positive results, which might hold more generally, should encourage the exploration of developing new MOEAs in the area.

Xinnian Liang, Shuangzhi Wu, Hui Huang et al.

Retrieval augmented methods have shown promising results in various classification tasks. However, existing methods focus on retrieving extra context to enrich the input, which is noise sensitive and non-expandable. In this paper, following this line, we propose a $k$-nearest-neighbor (KNN) -based method for retrieval augmented classifications, which interpolates the predicted label distribution with retrieved instances' label distributions. Different from the standard KNN process, we propose a decoupling mechanism as we find that shared representation for classification and retrieval hurts performance and leads to training instability. We evaluate our method on a wide range of classification datasets. Experimental results demonstrate the effectiveness and robustness of our proposed method. We also conduct extra experiments to analyze the contributions of different components in our model.\footnote{\url{https://github.com/xnliang98/knn-cls-w-decoupling}}

Xinwei Wu, Junzhuo Li, Minghui Xu et al.

Large language models pretrained on a huge amount of data capture rich knowledge and information in the training data. The ability of data memorization and regurgitation in pretrained language models, revealed in previous studies, brings the risk of data leakage. In order to effectively reduce these risks, we propose a framework DEPN to Detect and Edit Privacy Neurons in pretrained language models, partially inspired by knowledge neurons and model editing. In DEPN, we introduce a novel method, termed as privacy neuron detector, to locate neurons associated with private information, and then edit these detected privacy neurons by setting their activations to zero. Furthermore, we propose a privacy neuron aggregator dememorize private information in a batch processing manner. Experimental results show that our method can significantly and efficiently reduce the exposure of private data leakage without deteriorating the performance of the model. Additionally, we empirically demonstrate the relationship between model memorization and privacy neurons, from multiple perspectives, including model size, training time, prompts, privacy neuron distribution, illustrating the robustness of our approach.

Yu-Ran Gu, Chao Bian, Chao Qian

Submodular maximization arises in many applications, and has attracted a lot of research attentions from various areas such as artificial intelligence, finance and operations research. Previous studies mainly consider only one kind of constraint, while many real-world problems often involve several constraints. In this paper, we consider the problem of submodular maximization under the intersection of two commonly used constraints, i.e., $k$-matroid constraint and $m$-knapsack constraint, and propose a new algorithm SPROUT by incorporating partial enumeration into the simultaneous greedy framework. We prove that SPROUT can achieve a polynomial-time approximation guarantee better than the state-of-the-art algorithms. Then, we introduce the random enumeration and smooth techniques into SPROUT to improve its efficiency, resulting in the SPROUT++ algorithm, which can keep a similar approximation guarantee. Experiments on the applications of movie recommendation and weighted max-cut demonstrate the superiority of SPROUT++ in practice.

Weilong Dong, Xinwei Wu, Junzhuo Li et al.

Massively multi-task learning with large language models has recently made substantial progress on few-shot generalization. However, this is usually performed in a centralized learning fashion, ignoring the privacy sensitivity issue of (annotated) data used in multiple tasks. To mitigate this issue, we propose FewFedWeight, a few-shot federated learning framework across multiple tasks, to achieve the best of both worlds: privacy preservation and cross-task generalization. FewFedWeight trains client models in isolated devices without sharing data. It broadcasts the global model in the server to each client and produces pseudo data for clients so that knowledge from the global model can be explored to enhance few-shot learning of each client model. An energy-based algorithm is further proposed to weight pseudo samples in order to reduce the negative impact of noise from the generated pseudo data. Adaptive model weights of client models are also tuned according to their performance. We use these model weights to dynamically aggregate client models to update the global model. Experiments on 118 NLP tasks show that FewFedWeight can significantly improve the performance of client models on 61% tasks with an average performance improvement rate of 30.5% over the baseline and substantially outperform FedAvg and other decentralized learning methods.

Junzhuo Li, Xinwei Wu, Weilong Dong et al.

Knowledge distillation (KD) has been widely used for model compression and knowledge transfer. Typically, a big teacher model trained on sufficient data transfers knowledge to a small student model. However, despite the success of KD, little effort has been made to study whether KD leaks the training data of the teacher model. In this paper, we experimentally reveal that KD suffers from the risk of privacy leakage. To alleviate this issue, we propose a novel knowledge distillation method, swing distillation, which can effectively protect the private information of the teacher model from flowing to the student model. In our framework, the temperature coefficient is dynamically and adaptively adjusted according to the degree of private information contained in the data, rather than a predefined constant hyperparameter. It assigns different temperatures to tokens according to the likelihood that a token in a position contains private information. In addition, we inject noise into soft targets provided to the student model, in order to avoid unshielded knowledge transfer. Experiments on multiple datasets and tasks demonstrate that the proposed swing distillation can significantly reduce (by over 80% in terms of canary exposure) the risk of privacy leakage in comparison to KD with competitive or better performance. Furthermore, swing distillation is robust against the increasing privacy budget.

Yun Li, Zhe Liu, Yajing Kong et al.

Applying Multimodal Large Language Models (MLLMs) to video understanding presents significant challenges due to the need to model temporal relations across frames. Existing approaches adopt either implicit temporal modeling, relying solely on the LLM decoder, or explicit temporal modeling, employing auxiliary temporal encoders. To investigate this debate between the two paradigms, we propose the Stackable Temporal Encoder (STE). STE enables flexible explicit temporal modeling with adjustable temporal receptive fields and token compression ratios. Using STE, we systematically compare implicit and explicit temporal modeling across dimensions such as overall performance, token compression effectiveness, and temporal-specific understanding. We also explore STE's design considerations and broader impacts as a plug-in module and in image modalities. Our findings emphasize the critical role of explicit temporal modeling, providing actionable insights to advance video MLLMs.

Haitao Liu, Kai Wu, Yew-Soon Ong et al.

Multi-task Gaussian process (MTGP) is a well-known non-parametric Bayesian model for learning correlated tasks effectively by transferring knowledge across tasks. But current MTGPs are usually limited to the multi-task scenario defined in the same input domain, leaving no space for tackling the heterogeneous case, i.e., the features of input domains vary over tasks. To this end, this paper presents a novel heterogeneous stochastic variational linear model of coregionalization (HSVLMC) model for simultaneously learning the tasks with varied input domains. Particularly, we develop the stochastic variational framework with Bayesian calibration that (i) takes into account the effect of dimensionality reduction raised by domain mappings in order to achieve effective input alignment; and (ii) employs a residual modeling strategy to leverage the inductive bias brought by prior domain mappings for better model inference. Finally, the superiority of the proposed model against existing LMC models has been extensively verified on diverse heterogeneous multi-task cases and a practical multi-fidelity steam turbine exhaust problem.

Chao Bian, Chao Qian, Yang Yu et al.

Evolutionary algorithms (EAs) are a sort of nature-inspired metaheuristics, which have wide applications in various practical optimization problems. In these problems, objective evaluations are usually inaccurate, because noise is almost inevitable in real world, and it is a crucial issue to weaken the negative effect caused by noise. Sampling is a popular strategy, which evaluates the objective a couple of times, and employs the mean of these evaluation results as an estimate of the objective value. In this work, we introduce a novel sampling method, median sampling, into EAs, and illustrate its properties and usefulness theoretically by solving OneMax, the problem of maximizing the number of 1s in a bit string. Instead of the mean, median sampling employs the median of the evaluation results as an estimate. Through rigorous theoretical analysis on OneMax under the commonly used onebit noise, we show that median sampling reduces the expected runtime exponentially. Next, through two special noise models, we show that when the 2-quantile of the noisy fitness increases with the true fitness, median sampling can be better than mean sampling; otherwise, it may fail and mean sampling can be better. The results may guide us to employ median sampling properly in practical applications.

Chao Bian, Chao Qian, Ke Tang et al.

Evolutionary algorithms (EAs) have found many successful real-world applications, where the optimization problems are often subject to a wide range of uncertainties. To understand the practical behaviors of EAs theoretically, there are a series of efforts devoted to analyzing the running time of EAs for optimization under uncertainties. Existing studies mainly focus on noisy and dynamic optimization, while another common type of uncertain optimization, i.e., robust optimization, has been rarely touched. In this paper, we analyze the expected running time of the (1+1)-EA solving robust linear optimization problems (i.e., linear problems under robust scenarios) with a cardinality constraint $k$. Two common robust scenarios, i.e., deletion-robust and worst-case, are considered. Particularly, we derive tight ranges of the robust parameter $d$ or budget $k$ allowing the (1+1)-EA to find an optimal solution in polynomial running time, which disclose the potential of EAs for robust optimization.

Chao Qian, Chao Bian, Yang Yu et al.

In noisy evolutionary optimization, sampling is a common strategy to deal with noise. By the sampling strategy, the fitness of a solution is evaluated multiple times (called \emph{sample size}) independently, and its true fitness is then approximated by the average of these evaluations. Most previous studies on sampling are empirical, and the few theoretical studies mainly showed the effectiveness of sampling with a sufficiently large sample size. In this paper, we theoretically examine what strategies can work when sampling with any fixed sample size fails. By constructing a family of artificial noisy examples, we prove that sampling is always ineffective, while using parent or offspring populations can be helpful on some examples. We also construct an artificial noisy example to show that when using neither sampling nor populations is effective, a tailored adaptive sampling (i.e., sampling with an adaptive sample size) strategy can work. These findings may enhance our understanding of sampling to some extent, but future work is required to validate them in natural situations.

Chao Qian, Chao Bian, Wu Jiang et al.

In many real-world optimization problems, the objective function evaluation is subject to noise, and we cannot obtain the exact objective value. Evolutionary algorithms (EAs), a type of general-purpose randomized optimization algorithm, have been shown to be able to solve noisy optimization problems well. However, previous theoretical analyses of EAs mainly focused on noise-free optimization, which makes the theoretical understanding largely insufficient for the noisy case. Meanwhile, the few existing theoretical studies under noise often considered the one-bit noise model, which flips a randomly chosen bit of a solution before evaluation; while in many realistic applications, several bits of a solution can be changed simultaneously. In this paper, we study a natural extension of one-bit noise, the bit-wise noise model, which independently flips each bit of a solution with some probability. We analyze the running time of the (1+1)-EA solving OneMax and LeadingOnes under bit-wise noise for the first time, and derive the ranges of the noise level for polynomial and super-polynomial running time bounds. The analysis on LeadingOnes under bit-wise noise can be easily transferred to one-bit noise, and improves the previously known results. Since our analysis discloses that the (1+1)-EA can be efficient only under low noise levels, we also study whether the sampling strategy can bring robustness to noise. We prove that using sampling can significantly increase the largest noise level allowing a polynomial running time, that is, sampling is robust to noise.