Xinghua Zhang

Xinghua Zhang, Bowen Yu, Haiyang Yu et al.

Measuring the quality of responses generated by LLMs is a challenging task, particularly when it comes to evaluating whether the response is aligned with human preference. A novel approach involves using the LLM itself to make evaluation and stabilizing the results through multiple independent evaluations, similar to a single-layer narrow LLM network. This network consists of a fixed number of neurons, with each neuron being the same LLM. In this paper, we draw upon the extensive research on deep neural networks to explore whether deeper and wider networks can lead to fairer evaluations. Specifically, inspired by the observation that different neurons in a neural network are responsible for detecting different concepts, we first adaptively generate as many neuron roles as possible for each evaluation sample. Each perspective corresponds to the role of a specific LLM neuron in the first layer. In subsequent layers, we follow the idea that higher layers in deep networks are responsible for more comprehensive features, each layer receives representations from all neurons in the previous layer, integrating the locally learned evaluation information to obtain a more comprehensive evaluation result. Interestingly, this network design resembles the process of academic paper reviewing. To validate the effectiveness of our method, we construct the largest and most diverse English evaluation benchmark LLMEval$^2$ for LLM evaluators, comprising 15 tasks, 8 abilities, and 2,553 samples. Experimental results demonstrate that a wider network (involving many reviewers) with 2 layers (one round of discussion) performs the best, improving kappa correlation coefficient from 0.28 to 0.34. We also leverage WideDeep to aid in the assessment of Chinese LLMs, which has accelerated the evaluation time by 4.6 times, resulting in a 60% cost saving. WideDeep achieves a remarkable 93% agreement level among humans.

Gaode Chen, Yijun Su, Xinghua Zhang et al.

Data insufficiency problems (i.e., data missing and label scarcity) caused by inadequate services and infrastructures or imbalanced development levels of cities have seriously affected the urban computing tasks in real scenarios. Prior transfer learning methods inspire an elegant solution to the data insufficiency, but are only concerned with one kind of insufficiency issue and fail to give consideration to both sides. In addition, most previous cross-city transfer methods overlook inter-city data privacy which is a public concern in practical applications. To address the above challenging problems, we propose a novel Cross-city Federated Transfer Learning framework (CcFTL) to cope with the data insufficiency and privacy problems. Concretely, CcFTL transfers the relational knowledge from multiple rich-data source cities to the target city. Besides, the model parameters specific to the target task are firstly trained on the source data and then fine-tuned to the target city by parameter transfer. With our adaptation of federated training and homomorphic encryption settings, CcFTL can effectively deal with the data privacy problem among cities. We take the urban region profiling as an application of smart cities and evaluate the proposed method with a real-world study. The experiments demonstrate the notable superiority of our framework over several competitive state-of-the-art methods.

Xinghua Zhang, Haiyang Yu, Yongbin Li et al.

In the era of large language models (LLMs), a vast amount of conversation logs will be accumulated thanks to the rapid development trend of language UI. Conversation Analysis (CA) strives to uncover and analyze critical information from conversation data, streamlining manual processes and supporting business insights and decision-making. The need for CA to extract actionable insights and drive empowerment is becoming increasingly prominent and attracting widespread attention. However, the lack of a clear scope for CA leads to a dispersion of various techniques, making it difficult to form a systematic technical synergy to empower business applications. In this paper, we perform a thorough review and systematize CA task to summarize the existing related work. Specifically, we formally define CA task to confront the fragmented and chaotic landscape in this field, and derive four key steps of CA from conversation scene reconstruction, to in-depth attribution analysis, and then to performing targeted training, finally generating conversations based on the targeted training for achieving the specific goals. In addition, we showcase the relevant benchmarks, discuss potential challenges and point out future directions in both industry and academia. In view of current advancements, it is evident that the majority of efforts are still concentrated on the analysis of shallow conversation elements, which presents a considerable gap between the research and business, and with the assist of LLMs, recent work has shown a trend towards research on causality and strategic tasks which are sophisticated and high-level. The analyzed experiences and insights will inevitably have broader application value in business operations that target conversation logs.

Gehang Zhang, Bowen Yu, Jiangxia Cao et al.

Graph contrastive learning (GCL) has recently achieved substantial advancements. Existing GCL approaches compare two different ``views'' of the same graph in order to learn node/graph representations. The underlying assumption of these studies is that the graph augmentation strategy is capable of generating several different graph views such that the graph views are structurally different but semantically similar to the original graphs, and thus the ground-truth labels of the original and augmented graph/nodes can be regarded identical in contrastive learning. However, we observe that this assumption does not always hold. For instance, the deletion of a super-node within a social network can exert a substantial influence on the partitioning of communities for other nodes. Similarly, any perturbation to nodes or edges in a molecular graph will change the labels of the graph. Therefore, we believe that augmenting the graph, accompanied by an adaptation of the labels used for the contrastive loss, will facilitate the encoder to learn a better representation. Based on this idea, we propose ID-MixGCL, which allows the simultaneous interpolation of input nodes and corresponding identity labels to obtain soft-confidence samples, with a controllable degree of change, leading to the capture of fine-grained representations from self-supervised training on unlabeled graphs. Experimental results demonstrate that ID-MixGCL improves performance on graph classification and node classification tasks, as demonstrated by significant improvements on the Cora, IMDB-B, IMDB-M, and PROTEINS datasets compared to state-of-the-art techniques, by 3-29% absolute points.

Wenyuan Zhang, Shuaiyi Nie, Jiawei Sheng et al.

Large language model (LLM) role-playing has gained widespread attention. Authentic character knowledge is crucial for constructing realistic LLM role-playing agents. However, existing works usually overlook the exploration of LLMs' ability to detect characters' known knowledge errors (KKE) and unknown knowledge errors (UKE) while playing roles, which would lead to low-quality automatic construction of character trainable corpus. In this paper, we propose RoleKE-Bench to evaluate LLMs' ability to detect errors in KKE and UKE. The results indicate that even the latest LLMs struggle to detect these two types of errors effectively, especially when it comes to familiar knowledge. We experimented with various reasoning strategies and propose an agent-based reasoning method, Self-Recollection and Self-Doubt (S$^2$RD), to explore further the potential for improving error detection capabilities. Experiments show that our method effectively improves the LLMs' ability to detect error character knowledge, but it remains an issue that requires ongoing attention.

Taoyu Su, Xinghua Zhang, Jiawei Sheng et al.

Multi-modal entity alignment (MMEA) aims to identify equivalent entities between two multi-modal knowledge graphs (MMKGs), whose entities can be associated with relational triples and related images. Most previous studies treat the graph structure as a special modality, and fuse different modality information with separate uni-modal encoders, neglecting valuable relational associations in modalities. Other studies refine each uni-modal information with graph structures, but may introduce unnecessary relations in specific modalities. To this end, we propose a novel local-to-global interaction network for MMEA, termed as LoginMEA. Particularly, we first fuse local multi-modal interactions to generate holistic entity semantics and then refine them with global relational interactions of entity neighbors. In this design, the uni-modal information is fused adaptively, and can be refined with relations accordingly. To enrich local interactions of multi-modal entity information, we device modality weights and low-rank interactive fusion, allowing diverse impacts and element-level interactions among modalities. To capture global interactions of graph structures, we adopt relation reflection graph attention networks, which fully capture relational associations between entities. Extensive experiments demonstrate superior results of our method over 5 cross-KG or bilingual benchmark datasets, indicating the effectiveness of capturing local and global interactions.

Taoyu Su, Jiawei Sheng, Shicheng Wang et al.

Multi-modal entity alignment (MMEA) aims to identify equivalent entities between multi-modal knowledge graphs (MMKGs), where the entities can be associated with related images. Most existing studies integrate multi-modal information heavily relying on the automatically-learned fusion module, rarely suppressing the redundant information for MMEA explicitly. To this end, we explore variational information bottleneck for multi-modal entity alignment (IBMEA), which emphasizes the alignment-relevant information and suppresses the alignment-irrelevant information in generating entity representations. Specifically, we devise multi-modal variational encoders to generate modal-specific entity representations as probability distributions. Then, we propose four modal-specific information bottleneck regularizers, limiting the misleading clues in refining modal-specific entity representations. Finally, we propose a modal-hybrid information contrastive regularizer to integrate all the refined modal-specific representations, enhancing the entity similarity between MMKGs to achieve MMEA. We conduct extensive experiments on two cross-KG and three bilingual MMEA datasets. Experimental results demonstrate that our model consistently outperforms previous state-of-the-art methods, and also shows promising and robust performance in low-resource and high-noise data scenarios.

Wenyuan Zhang, Xinghua Zhang, Haiyang Yu et al.

Experience intervention in web agents emerges as a promising technical paradigm, enhancing agent interaction capabilities by providing valuable insights from accumulated experiences. However, existing methods predominantly inject experience passively as global context before task execution, struggling to adapt to dynamically changing contextual observations during agent-environment interaction. We propose ExpSeek, which shifts experience toward step-level proactive seeking: (1) estimating step-level entropy thresholds to determine intervention timing using the model's intrinsic signals; (2) designing step-level tailor-designed experience content. Experiments on Qwen3-8B and 32B models across four challenging web agent benchmarks demonstrate that ExpSeek achieves absolute improvements of 9.3% and 7.5%, respectively. Our experiments validate the feasibility and advantages of entropy as a self-triggering signal, reveal that even a 4B small-scale experience model can significantly boost the performance of larger agent models.

Zhenhong Zhou, Haiyang Yu, Xinghua Zhang et al.

Large language models (LLMs) rely on safety alignment to avoid responding to malicious user inputs. Unfortunately, jailbreak can circumvent safety guardrails, resulting in LLMs generating harmful content and raising concerns about LLM safety. Due to language models with intensive parameters often regarded as black boxes, the mechanisms of alignment and jailbreak are challenging to elucidate. In this paper, we employ weak classifiers to explain LLM safety through the intermediate hidden states. We first confirm that LLMs learn ethical concepts during pre-training rather than alignment and can identify malicious and normal inputs in the early layers. Alignment actually associates the early concepts with emotion guesses in the middle layers and then refines them to the specific reject tokens for safe generations. Jailbreak disturbs the transformation of early unethical classification into negative emotions. We conduct experiments on models from 7B to 70B across various model families to prove our conclusion. Overall, our paper indicates the intrinsical mechanism of LLM safety and how jailbreaks circumvent safety guardrails, offering a new perspective on LLM safety and reducing concerns. Our code is available at https://github.com/ydyjya/LLM-IHS-Explanation.

Zhenhong Zhou, Haiyang Yu, Xinghua Zhang et al.

Large language models (LLMs) achieve state-of-the-art performance on multiple language tasks, yet their safety guardrails can be circumvented, leading to harmful generations. In light of this, recent research on safety mechanisms has emerged, revealing that when safety representations or component are suppressed, the safety capability of LLMs are compromised. However, existing research tends to overlook the safety impact of multi-head attention mechanisms, despite their crucial role in various model functionalities. Hence, in this paper, we aim to explore the connection between standard attention mechanisms and safety capability to fill this gap in the safety-related mechanistic interpretability. We propose a novel metric which tailored for multi-head attention, the Safety Head ImPortant Score (Ships), to assess the individual heads' contributions to model safety. Based on this, we generalize Ships to the dataset level and further introduce the Safety Attention Head AttRibution Algorithm (Sahara) to attribute the critical safety attention heads inside the model. Our findings show that the special attention head has a significant impact on safety. Ablating a single safety head allows aligned model (e.g., Llama-2-7b-chat) to respond to 16 times more harmful queries, while only modifying 0.006% of the parameters, in contrast to the ~ 5% modification required in previous studies. More importantly, we demonstrate that attention heads primarily function as feature extractors for safety and models fine-tuned from the same base model exhibit overlapping safety heads through comprehensive experiments. Together, our attribution approach and findings provide a novel perspective for unpacking the black box of safety mechanisms within large models.

Xinghua Zhang, Haiyang Yu, Cheng Fu et al.

In the realm of large language models (LLMs), the ability of models to accurately follow instructions is paramount as more agents and applications leverage LLMs for construction, where the complexity of instructions are rapidly increasing. However, on the one hand, there is only a certain amount of complex instruction evaluation data; on the other hand, there are no dedicated algorithms to improve the ability to follow complex instructions. To this end, this paper introduces TRACE, a benchmark for improving and evaluating the complex instructionfollowing ability, which consists of 120K training data and 1K evaluation data. Furthermore, we propose IOPO (Input-Output Preference Optimization) alignment method which takes both input and output preference pairs into consideration, where LLMs not only rapidly align with response preferences but also meticulously explore the instruction preferences. Extensive experiments on both in-domain and outof-domain datasets confirm the effectiveness of IOPO, showing 8.15%, 2.18% improvements on in-domain data and 6.29%, 3.13% on outof-domain data compared to SFT and DPO respectively.

Wenyuan Zhang, Shuaiyi Nie, Xinghua Zhang et al.

We introduce S1-Bench, a novel benchmark designed to evaluate the performance of Large Reasoning Models (LRMs) on simple tasks that favor intuitive system 1 thinking rather than deliberative system 2 reasoning. While LRMs have achieved significant breakthroughs in complex reasoning tasks through explicit chains of thought, their heavy reliance on system 2 thinking may limit their system 1 thinking capabilities. However, there is a lack of an appropriate benchmark for evaluating LRM's system 1 thinking capabilities. To fill this gap, S1-Bench introduces a suite of simple, diverse, and natural questions across multiple domains and languages, specifically designed to assess LRMs' performance on questions more suitable for system 1 . We conduct extensive evaluations across 28 LRMs, revealing their inefficiency, inadequate accuracy, and limited robustness when handling simple questions. Additionally, we observe a gap between their difficulty perception and generation length. Overall, this work paves the way toward dual-system compatibility in the development of LRMs.

Xiaodong Li, Jiawei Sheng, Jiangxia Cao et al.

Cross-domain recommendation (CDR) has demonstrated to be an effective solution for alleviating the user cold-start issue. By leveraging rich user-item interactions available in a richly informative source domain, CDR could improve the recommendation performance for cold-start users in the target domain. Previous CDR approaches mostly adhere the Embedding and Mapping (EMCDR) paradigm, which learns a user-shared mapping function to transfer users' preference from the source domain to the target domain, neglecting users' personalized preference. Recent CDR approaches further leverage the meta-learning paradigm, considering the CDR task for each user independently and learning user-specific mapping functions for each user. However, they mostly learn representations for each user individually, which ignores the common preference between different users, neglecting valuable information for CDR. In addition, all these approaches usually summarize the user's preference into an overall representation, which can hardly capture the user's multi-interest preference. To this end, we propose a personalized multi-interest modeling framework for CDR to cold-start users, termed as NF-NPCDR. Specifically, we propose a personalized preference encoder that enhances the neural process (NP) with the normalizing flow (NF) to convert the Gaussian (unimodal) distribution to a multimodal distribution, providing a novel way to capture the user's personalized multi-interest preference. Then, we propose a common preference encoder with a preference pool to capture the common preference between different users. Furthermore, we introduce a stochastic adaptive decoder to incorporate both the personalized and common preference for cold-start users, adaptively modulating both preference for better recommendation.

Wenyuan Zhang, Xinghua Zhang, Shiyao Cui et al.

Aspect sentiment quad prediction (ASQP) aims to predict the quad sentiment elements for a given sentence, which is a critical task in the field of aspect-based sentiment analysis. However, the data imbalance issue has not received sufficient attention in ASQP task. In this paper, we divide the issue into two-folds, quad-pattern imbalance and aspect-category imbalance, and propose an Adaptive Data Augmentation (ADA) framework to tackle the imbalance issue. Specifically, a data augmentation process with a condition function adaptively enhances the tail quad patterns and aspect categories, alleviating the data imbalance in ASQP. Following previous studies, we also further explore the generative framework for extracting complete quads by introducing the category prior knowledge and syntax-guided decoding target. Experimental results demonstrate that data augmentation for imbalance in ASQP task can improve the performance, and the proposed ADA method is superior to naive data oversampling.

Minzheng Wang, Yongbin Li, Haobo Wang et al.

Effective social intelligence simulation requires language agents to dynamically adjust reasoning depth, a capability notably absent in current studies. Existing methods either lack this kind of reasoning capability or enforce Long Chain-of-Thought reasoning uniformly across all scenarios, resulting in excessive token usage and inflexible social simulation. To address this, we propose an $\textbf{A}$daptive $\textbf{M}$ode $\textbf{L}$earning ($\textbf{AML}$) framework in this paper, aiming to improve the adaptive thinking ability of language agents in dynamic social interactions. To this end, we first identify hierarchical thinking modes ranging from intuitive response to deep deliberation based on the cognitive control theory. We then develop the $\textbf{A}$daptive $\textbf{M}$ode $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{AMPO}$) algorithm to optimize the context-aware mode switching and reasoning. Our framework advances existing research in three key aspects: (1) Multi-granular thinking mode design, (2) Context-aware mode switching across social interaction, and (3) Token-efficient reasoning via depth-adaptive processing. Extensive experiments on social intelligence benchmarks verify that AML achieves 15.6% higher task performance than GPT-4o. Notably, our AMPO outperforms GRPO by 7.0% with 32.8% shorter reasoning chains, demonstrating the advantage of adaptive thinking mode selection and optimization mechanism in AMPO over GRPO's fixed-depth solution.

Minzheng Wang, Xinghua Zhang, Kun Chen et al.

Large language models (LLMs) enabled dialogue systems have become one of the central modes in human-machine interaction, which bring about vast amounts of conversation logs and increasing demand for dialogue generation. The dialogue's life-cycle spans from $\textit{Prelude}$ through $\textit{Interlocution}$ to $\textit{Epilogue}$, encompassing rich dialogue elements. Despite large volumes of dialogue-related studies, there is a lack of systematic investigation into the dialogue stages to frame benchmark construction that covers comprehensive dialogue elements. This hinders the precise modeling, generation and assessment of LLMs-based dialogue systems. To bridge this gap, in this paper, we introduce a new research task--$\textbf{D}$ialogue $\textbf{E}$lement $\textbf{MO}$deling, including $\textit{Element Awareness}$ and $\textit{Dialogue Agent Interaction}$, and propose a novel benchmark, $\textbf{DEMO}$, designed for a comprehensive dialogue modeling and assessment. On this basis, we further build the DEMO agent with the adept ability to model dialogue elements via imitation learning. Extensive experiments on DEMO indicate that current representative LLMs still have considerable potential for enhancement, and our DEMO agent performs well in both dialogue element modeling and out-of-domain tasks.

Shipeng Li, Shikun Li, Zhiqin Yang et al.

Reinforcement learning (RL) has become a key technique for enhancing LLMs' reasoning abilities, yet its data inefficiency remains a major bottleneck. To address this critical yet challenging issue, we present a novel gradient-alignment-based method, named LearnAlign, which intelligently selects the learnable and representative training reasoning data for RL post-training. To overcome the issue of response-length bias in gradient norms, we introduce the data learnability based on the success rate, which can indicate the learning potential of each data point. Experiments across three mathematical reasoning benchmarks demonstrate that our method significantly reduces training data requirements while achieving minor performance degradation or even improving performance compared to full-data training. For example, it reduces data requirements by up to 1,000 data points with better performance (77.53%) than that on the full dataset on GSM8K benchmark (77.04%). Furthermore, we show its effectiveness in the staged RL setting. This work provides valuable insights into data-efficient RL post-training and establishes a foundation for future research in optimizing reasoning data selection. To facilitate future work, we will release code.

Xiang Li, Haiyang Yu, Xinghua Zhang et al.

Process Reward Models (PRMs) are crucial in complex reasoning and problem-solving tasks (e.g., LLM agents with long-horizon decision-making) by verifying the correctness of each intermediate reasoning step. In real-world scenarios, LLMs may apply various reasoning patterns (e.g., decomposition) to solve a problem, potentially suffering from errors under various reasoning patterns. Therefore, PRMs are required to identify errors under various reasoning patterns during the reasoning process. However, existing benchmarks mainly focus on evaluating PRMs with stepwise correctness, ignoring a systematic evaluation of PRMs under various reasoning patterns. To mitigate this gap, we introduce Socratic-PRMBench, a new benchmark to evaluate PRMs systematically under six reasoning patterns, including Transformation, Decomposition, Regather, Deduction, Verification, and Integration. Socratic-PRMBench}comprises 2995 reasoning paths with flaws within the aforementioned six reasoning patterns. Through our experiments on both PRMs and LLMs prompted as critic models, we identify notable deficiencies in existing PRMs. These observations underscore the significant weakness of current PRMs in conducting evaluations on reasoning steps under various reasoning patterns. We hope Socratic-PRMBench can serve as a comprehensive testbed for systematic evaluation of PRMs under diverse reasoning patterns and pave the way for future development of PRMs.

Tao Zou, Xinghua Zhang, Haiyang Yu et al.

With the development and widespread application of large language models (LLMs), the new paradigm of "Model as Product" is rapidly evolving, and demands higher capabilities to address complex user needs, often requiring precise workflow execution which involves the accurate understanding of multiple tasks. However, existing benchmarks focusing on single-task environments with limited constraints lack the complexity required to fully reflect real-world scenarios. To bridge this gap, we present the Extremely Complex Instruction Following Benchmark (EIFBENCH), meticulously crafted to facilitate a more realistic and robust evaluation of LLMs. EIFBENCH not only includes multi-task scenarios that enable comprehensive assessment across diverse task types concurrently, but also integrates a variety of constraints, replicating complex operational environments. Furthermore, we propose the Segment Policy Optimization (SegPO) algorithm to enhance the LLM's ability to accurately fulfill multi-task workflow. Evaluations on EIFBENCH have unveiled considerable performance discrepancies in existing LLMs when challenged with these extremely complex instructions. This finding underscores the necessity for ongoing optimization to navigate the intricate challenges posed by LLM applications.

Minzheng Wang, Longze Chen, Cheng Fu et al.

Long-context modeling capabilities have garnered widespread attention, leading to the emergence of Large Language Models (LLMs) with ultra-context windows. Meanwhile, benchmarks for evaluating long-context LLMs are gradually catching up. However, existing benchmarks employ irrelevant noise texts to artificially extend the length of test cases, diverging from the real-world scenarios of long-context applications. To bridge this gap, we propose a novel long-context benchmark, Loong, aligning with realistic scenarios through extended multi-document question answering (QA). Unlike typical document QA, in Loong's test cases, each document is relevant to the final answer, ignoring any document will lead to the failure of the answer. Furthermore, Loong introduces four types of tasks with a range of context lengths: Spotlight Locating, Comparison, Clustering, and Chain of Reasoning, to facilitate a more realistic and comprehensive evaluation of long-context understanding. Extensive experiments indicate that existing long-context language models still exhibit considerable potential for enhancement. Retrieval augmented generation (RAG) achieves poor performance, demonstrating that Loong can reliably assess the model's long-context modeling capabilities.

Xinghua Zhang, Bowen Yu, Tingwen Liu et al.

Distantly supervised named entity recognition (DS-NER) efficiently reduces labor costs but meanwhile intrinsically suffers from the label noise due to the strong assumption of distant supervision. Typically, the wrongly labeled instances comprise numbers of incomplete and inaccurate annotation noise, while most prior denoising works are only concerned with one kind of noise and fail to fully explore useful information in the whole training set. To address this issue, we propose a robust learning paradigm named Self-Collaborative Denoising Learning (SCDL), which jointly trains two teacher-student networks in a mutually-beneficial manner to iteratively perform noisy label refinery. Each network is designed to exploit reliable labels via self denoising, and two networks communicate with each other to explore unreliable annotations by collaborative denoising. Extensive experimental results on five real-world datasets demonstrate that SCDL is superior to state-of-the-art DS-NER denoising methods.

Gaode Chen, Xinghua Zhang, Yanyan Zhao et al.

Sequential recommendation systems alleviate the problem of information overload, and have attracted increasing attention in the literature. Most prior works usually obtain an overall representation based on the user's behavior sequence, which can not sufficiently reflect the multiple interests of the user. To this end, we propose a novel method called PIMI to mitigate this issue. PIMI can model the user's multi-interest representation effectively by considering both the periodicity and interactivity in the item sequence. Specifically, we design a periodicity-aware module to utilize the time interval information between user's behaviors. Meanwhile, an ingenious graph is proposed to enhance the interactivity between items in user's behavior sequence, which can capture both global and local item features. Finally, a multi-interest extraction module is applied to describe user's multiple interests based on the obtained item representation. Extensive experiments on two real-world datasets Amazon and Taobao show that PIMI outperforms state-of-the-art methods consistently.