Zhuosheng Zhang

Recent Publications [Show All]

• Caution for the Environment: Multimodal Agents are Susceptible to Environmental Distractions
  Xinbei Ma, Yiting Wang, Yao Yao, Tongxin Yuan, Aston Zhang, Zhuosheng Zhang*, Hai Zhao*.
  arXiv, 2024
  [PDF] [Abstract] [Bib]
  

  This paper investigates the faithfulness of multimodal large language model (MLLM) agents in the graphical user interface (GUI) environment, aiming to address the research question of whether multimodal GUI agents can be distracted by environmental context. A general setting is proposed where both the user and the agent are benign, and the environment, while not malicious, contains unrelated content. A wide range of MLLMs are evaluated as GUI agents using our simulated dataset, following three working patterns with different levels of perception. Experimental results reveal that even the most powerful models, whether generalist agents or specialist GUI agents, are susceptible to distractions. While recent studies predominantly focus on the helpfulness (i.e., action accuracy) of multimodal agents, our findings indicate that these agents are prone to environmental distractions, resulting in unfaithful behaviors. Furthermore, we switch to the adversarial perspective and implement environment injection, demonstrating that such unfaithfulness can be exploited, leading to unexpected risks.

  @article{ma2024caution,
    title={Caution for the Environment: Multimodal Agents are Susceptible to Environmental Distractions},
    author={Ma, Xinbei and Wang, Yiting and Yao, Yao and Yuan, Tongxin and Zhang, Aston and Zhang, Zhuosheng and Zhao, Hai},
    journal={arXiv preprint arXiv:2408.02544},
    year={2024}
  }
                

• Flooding Spread of Manipulated Knowledge in LLM-Based Multi-Agent Communities
  Tianjie Ju, Yiting Wang, Xinbei Ma, Pengzhou Cheng, Haodong Zhao, Yulong Wang, Lifeng Liu, Jian Xie, Zhuosheng Zhang*, Gongshen Liu*.
  arXiv, 2024
  [PDF] [Abstract] [Bib]
  

  The rapid adoption of large language models (LLMs) in multi-agent systems has highlighted their impressive capabilities in various applications, such as collaborative problem-solving and autonomous negotiation. However, the security implications of these LLM-based multi-agent systems have not been thoroughly investigated, particularly concerning the spread of manipulated knowledge. In this paper, we investigate this critical issue by constructing a detailed threat model and a comprehensive simulation environment that mirrors real-world multi-agent deployments in a trusted platform. Subsequently, we propose a novel two-stage attack method involving Persuasiveness Injection and Manipulated Knowledge Injection to systematically explore the potential for manipulated knowledge (i.e., counterfactual and toxic knowledge) spread without explicit prompt manipulation. Our method leverages the inherent vulnerabilities of LLMs in handling world knowledge, which can be exploited by attackers to unconsciously spread fabricated information. Through extensive experiments, we demonstrate that our attack method can successfully induce LLM-based agents to spread both counterfactual and toxic knowledge without degrading their foundational capabilities during agent communication. Furthermore, we show that these manipulations can persist through popular retrieval-augmented generation frameworks, where several benign agents store and retrieve manipulated chat histories for future interactions. This persistence indicates that even after the interaction has ended, the benign agents may continue to be influenced by manipulated knowledge. Our findings reveal significant security risks in LLM-based multi-agent systems, emphasizing the imperative need for robust defenses against manipulated knowledge spread, such as introducing ``guardian'' agents and advanced fact-checking tools.

  @article{ju2024flooding,
    title={Flooding Spread of Manipulated Knowledge in LLM-Based Multi-Agent Communities},
    author={Ju, Tianjie and Wang, Yiting and Ma, Xinbei and Cheng, Pengzhou and Zhao, Haodong and Wang, Yulong and Liu, Lifeng and Xie, Jian and Zhang, Zhuosheng and Liu, Gongshen},
    journal={arXiv preprint arXiv:2407.07791},
    year={2024}
  }
                

• TrojanRAG: Retrieval-Augmented Generation Can Be Backdoor Driver in Large Language Models
  Pengzhou Cheng, Yidong Ding, Tianjie Ju, Zongru Wu, Wei Du, Ping Yi, Zhuosheng Zhang, Gongshen Liu.
  arXiv, 2024
  [PDF] [Abstract] [Bib]
  

  Large language models (LLMs) have raised concerns about potential security threats despite performing significantly in Natural Language Processing (NLP). Backdoor attacks initially verified that LLM is doing substantial harm at all stages, but the cost and robustness have been criticized. Attacking LLMs is inherently risky in security review, while prohibitively expensive. Besides, the continuous iteration of LLMs will degrade the robustness of backdoors. In this paper, we propose TrojanRAG, which employs a joint backdoor attack in the Retrieval-Augmented Generation, thereby manipulating LLMs in universal attack scenarios. Specifically, the adversary constructs elaborate target contexts and trigger sets. Multiple pairs of backdoor shortcuts are orthogonally optimized by contrastive learning, thus constraining the triggering conditions to a parameter subspace to improve the matching. To improve the recall of the RAG for the target contexts, we introduce a knowledge graph to construct structured data to achieve hard matching at a fine-grained level. Moreover, we normalize the backdoor scenarios in LLMs to analyze the real harm caused by backdoors from both attackers' and users' perspectives and further verify whether the context is a favorable tool for jailbreaking models. Extensive experimental results on truthfulness, language understanding, and harmfulness show that TrojanRAG exhibits versatility threats while maintaining retrieval capabilities on normal queries.

  @article{cheng2024trojanrag,
    title={TrojanRAG: Retrieval-Augmented Generation Can Be Backdoor Driver in Large Language Models},
    author={Cheng, Pengzhou and Ding, Yidong and Ju, Tianjie and Wu, Zongru and Du, Wei and Yi, Ping and Zhang, Zhuosheng and Liu, Gongshen},
    journal={arXiv preprint arXiv:2405.13401},
    year={2024}
  }
                

• Prioritizing Safeguarding Over Autonomy: Risks of LLM Agents for Science
  Xiangru Tang, Qiao Jin, Kunlun Zhu, Tongxin Yuan, Yichi Zhang, Wangchunshu Zhou, Meng Qu, Yilun Zhao, Jian Tang, Zhuosheng Zhang, Arman Cohan, Zhiyong Lu, Mark Gerstein.
  arXiv, 2024
  [PDF] [Abstract] [Bib]
  

  Intelligent agents powered by large language models (LLMs) have demonstrated substantial promise in autonomously conducting experiments and facilitating scientific discoveries across various disciplines. While their capabilities are promising, they also introduce novel vulnerabilities that demand careful consideration for safety. However, there exists a notable gap in the literature, as there has been no comprehensive exploration of these vulnerabilities. This position paper fills this gap by conducting a thorough examination of vulnerabilities in LLM-based agents within scientific domains, shedding light on potential risks associated with their misuse and emphasizing the need for safety measures. We begin by providing a comprehensive overview of the potential risks inherent to scientific LLM agents, taking into account user intent, the specific scientific domain, and their potential impact on the external environment. Then, we delve into the origins of these vulnerabilities and provide a scoping review of the limited existing works. Based on our analysis, we propose a triadic framework involving human regulation, agent alignment, and an understanding of environmental feedback (agent regulation) to mitigate these identified risks. Furthermore, we highlight the limitations and challenges associated with safeguarding scientific agents and advocate for the development of improved models, robust benchmarks, and comprehensive regulations to address these issues effectively.

  @article{tang2024prioritizing,
    title={Prioritizing Safeguarding Over Autonomy: Risks of LLM Agents for Science},
    author={Tang, Xiangru and Jin, Qiao Jin and Zhu, Kunlun and Yuan, Tongxin and Zhang, Yichi and Zhou, Wangchunshu and Qu, Meng and Zhao, Yilun and Tang, Jian and Zhang, Zhuosheng and Cohan, Arman and Lu, Zhiyong and Gerstein, Mark},
    journal={arXiv preprint arXiv:2402.04247},
    year={2024}
  }
                

• Igniting Language Intelligence: The Hitchhiker's Guide From Chain-of-Thought Reasoning to Language Agents
  Zhuosheng Zhang#, Yao Yao#, Aston Zhang, Xiangru Tang, Xinbei Ma, Zhiwei He, Yiming Wang, Mark Gerstein, Rui Wang, Gongshen Liu, Hai Zhao.
  arXiv, 2023
  "Join us on an exciting journey from chain-of-thought reasoning to language agent!"
  [PDF] [Abstract] [Bib]
  
  CoT-Igniting-Agent

  Large language models (LLMs) have dramatically enhanced the field of language intelligence, as demonstrably evidenced by their formidable empirical performance across a spectrum of complex reasoning tasks. Additionally, theoretical proofs have illuminated their emergent reasoning capabilities, providing a compelling showcase of their advanced cognitive abilities in linguistic contexts. Critical to their remarkable efficacy in handling complex reasoning tasks, LLMs leverage the intriguing chain-of-thought (CoT) reasoning techniques, obliging them to formulate intermediate steps en route to deriving an answer. The CoT reasoning approach has not only exhibited proficiency in amplifying reasoning performance but also in enhancing interpretability, controllability, and flexibility. In light of these merits, recent research endeavors have extended CoT reasoning methodologies to nurture the development of autonomous language agents, which adeptly adhere to language instructions and execute actions within varied environments. This survey paper orchestrates a thorough discourse, penetrating vital research dimensions, encompassing: (i) the foundational mechanics of CoT techniques, with a focus on elucidating the circumstances and justification behind its efficacy; (ii) the paradigm shift in CoT; and (iii) the burgeoning of language agents fortified by CoT approaches. Prospective research avenues envelop explorations into generalization, efficiency, customization, scaling, and safety. We hope to offer readers a comprehensive understanding of prevalent research areas such as CoT reasoning and language agents and illuminate the interconnections weaving through these areas. This paper caters to a wide audience, including beginners seeking comprehensive knowledge of CoT reasoning and language agents, as well as experienced researchers interested in foundational mechanics and engaging in cutting-edge discussions on these topics. A repository for the related papers is available at https://github.com/Zoeyyao27/CoT-Igniting-Agent.

  @article{zhang2023igniting,
    title={Igniting Language Intelligence: The Hitchhiker's Guide From Chain-of-Thought Reasoning to Language Agents},
    author={Zhang, Zhuosheng and Yao, Yao and Zhang, Aston and Tang, Xiangru and Ma, Xinbei and He, Zhiwei and Wang, Yiming and Gerstein, Mark and Wang, Rui and Liu, Gongshen and others},
    journal={arXiv preprint arXiv:2311.11797},
    year={2023}
  }
                

• Look before You Leap: Enhancing Attention and Vigilance regarding Harmful Content with GuidelineLLM
  Shaoqing Zhang, Zhuosheng Zhang, Kehai Chen, Rongxiang Weng, Muyun Yang, Tiejun Zhao, Min Zhang.
  AAAI, 2025
  [PDF] [Abstract] [Bib]
  
  GuidelineLLM

  Despite being empowered with alignment mechanisms, large language models (LLMs) are increasingly vulnerable to emerging jailbreak attacks that can compromise their alignment mechanisms. This vulnerability poses significant risks to the real-world applications. Existing work faces challenges in both training efficiency and generalization capabilities (i.e., Reinforcement Learning from Human Feedback and Red-Teaming). Developing effective strategies to enable LLMs to resist continuously evolving jailbreak attempts represents a significant challenge. To address this challenge, we propose a novel defensive paradigm called GuidelineLLM, which assists LLMs in recognizing queries that may have harmful content. Before LLMs respond to a query, GuidelineLLM first identifies potential risks associated with the query, summarizes these risks into guideline suggestions, and then feeds these guidelines to the responding LLMs. Importantly, our approach eliminates the necessity for additional safety fine-tuning of the LLMs themselves; only the GuidelineLLM requires fine-tuning. This characteristic enhances the general applicability of GuidelineLLM across various LLMs. Experimental results demonstrate that GuidelineLLM can significantly reduce the attack success rate (ASR) against the LLMs (an average reduction of 34.17\% ASR) while maintaining the helpfulness of the LLMs in handling benign queries. Code is available at https://github.com/sqzhang-lazy/GuidelineLLM.

  @article{zhang2024look,
    title={Look Before You Leap: Enhancing Attention and Vigilance Regarding Harmful Content with GuidelineLLM},
    author={Zhang, Shaoqing and Zhang, Zhuosheng and Chen, Kehai and Weng, Rongxiang and Yang, Muyun and Zhao, Tiejun and Zhang, Min},
    journal={arXiv preprint arXiv:2412.10423},
    year={2024}
  }
                

• Gracefully Filtering Backdoor Samples for Generative Large Language Models without Retraining
  Zongru Wu, Pengzhou Cheng, Lingyong Fang, Zhuosheng Zhang*, Gongshen Liu*.
  COLING, 2025
  [PDF] [Abstract] [Bib]
  
  GraceFul

  Backdoor attacks remain significant security threats to generative large language models (LLMs). Since generative LLMs output sequences of high-dimensional token logits instead of low-dimensional classification logits, most existing backdoor defense methods designed for discriminative models like BERT are ineffective for generative LLMs. Inspired by the observed differences in learning behavior between backdoor and clean mapping in the frequency space, we transform gradients of each training sample, directly influencing parameter updates, into the frequency space. Our findings reveal a distinct separation between the gradients of backdoor and clean samples in the frequency space. Based on this phenomenon, we propose Gradient Clustering in the Frequency Space for Backdoor Sample Filtering (GraCeFul), which leverages sample-wise gradients in the frequency space to effectively identify backdoor samples without requiring retraining LLMs. Experimental results show that GraCeFul outperforms baselines significantly. Notably, GraCeFul exhibits remarkable computational efficiency, achieving nearly 100% recall and F1 scores in identifying backdoor samples, reducing the average success rate of various backdoor attacks to 0% with negligible drops in clean accuracy across multiple free-style question answering datasets. Additionally, GraCeFul generalizes to Llama-2 and Vicuna. The codes are publicly available at https://github.com/ZrW00/GraceFul.

  @inproceedings{wu2024gracefully,
    title   = {Gracefully Filtering Backdoor Samples for Generative Large Language Models without Retraining},
    author  = {Wu, Zongru and Cheng, Pengzhou and Fang, Lingyong and Zhang, Zhuosheng and Liu, Gongshen},
    booktitle = {Proceedings of the 31st International Conference on Computational Linguistics (COLING 2025)},
    year    = {2025}
  }
                

• Trajectory Volatility for Out-of-Distribution Detection in Mathematical Reasoning
  Yiming Wang, Pei Zhang, Baosong Yang, Derek F. Wong, Zhuosheng Zhang, Rui Wang.
  NeurIPS, 2024
  [PDF] [Abstract] [Bib]
  
  OOD-Math-Reasoning

  Real-world data deviating from the independent and identically distributed (i.i.d.) assumption of in-distribution training data poses security threats to deep networks, thus advancing out-of-distribution (OOD) detection algorithms. Detection methods in generative language models (GLMs) mainly focus on uncertainty estimation and embedding distance measurement, with the latter proven to be most effective in traditional linguistic tasks like summarization and translation. However, another complex generative scenario mathematical reasoning poses significant challenges to embedding-based methods due to its high-density feature of output spaces, but this feature causes larger discrepancies in the embedding shift trajectory between different samples in latent spaces. Hence, we propose a trajectory-based method TV score, which uses trajectory volatility for OOD detection in mathematical reasoning. Experiments show that our method outperforms all traditional algorithms on GLMs under mathematical reasoning scenarios and can be extended to more applications with high-density features in output spaces, such as multiple-choice questions.

  @article{wang2024trajectory,
    title={Trajectory Volatility for Out-of-Distribution Detection in Mathematical Reasoning},
    author={Wang, Yiming and Zhang, Pei and Yang, Baosong and Wong, Derek F and Zhang, Zhuosheng and Wang, Rui},
    journal={arXiv preprint arXiv:2405.14039},
    year={2024}
  }
                

• Is it Possible to Edit Large Language Models Robustly?
  Xinbei Ma, Tianjie Ju, Jiyang Qiu, Zhuosheng Zhang*, Hai Zhao*, Lifeng Liu, Yulong Wang.
  EMNLP, 2024
  "The robustness of model editing remains an open question."
  [PDF] [Abstract] [Bib]
  
  edit_analysis

  Large language models (LLMs) have played a pivotal role in building communicative AI to imitate human behaviors but face the challenge of efficient customization. To tackle this challenge, recent studies have delved into the realm of model editing, which manipulates specific memories of language models and changes the related language generation. However, the robustness of model editing remains an open question. This work seeks to understand the strengths and limitations of editing methods, thus facilitating robust, realistic applications of communicative AI. Concretely, we conduct extensive analysis to address the three key research questions. Q1: Can edited LLMs behave consistently resembling communicative AI in realistic situations? Q2: To what extent does the rephrasing of prompts lead LLMs to deviate from the edited knowledge memory? Q3: Which knowledge features are correlated with the performance and robustness of editing? Our experimental results uncover a substantial disparity between existing editing methods and the practical application of LLMs. On rephrased prompts that are complex and flexible but common in realistic applications, the performance of editing experiences a significant decline. Further analysis shows that more popular knowledge is memorized better, easier to recall, and more challenging to edit effectively.

  @article{ma2024possible,
    title={Is it Possible to Edit Large Language Models Robustly?},
    author={Ma, Xinbei and Ju, Tianjie and Qiu, Jiyang and Zhang, Zhuosheng and Zhao, Hai and Liu, Lifeng and Wang, Yulong},
    journal={arXiv preprint arXiv:2402.05827},
    year={2024}
  }
                

• GLaPE: Gold Label-agnostic Prompt Evaluation and Optimization for Large Language Model
  Xuanchang Zhang, Zhuosheng Zhang*, Hai Zhao*.
  EMNLP, 2024
  [PDF] [Abstract] [Bib]
  
  GLaPE

  Despite the rapid progress of large language models (LLMs), their task performance remains sensitive to prompt design. Recent studies have explored leveraging the LLM itself as an optimizer to identify optimal prompts that maximize task accuracy. However, when evaluating prompts, such approaches heavily rely on elusive manually annotated gold labels to calculate task accuracy for each candidate prompt, which hinders the widespread implementation and generality. To overcome the limitation, this work proposes a gold label-agnostic prompt evaluation (GLaPE) to alleviate dependence on gold labels. Motivated by the observed correlation between self-consistency and the accuracy of the answer, we adopt self-consistency as the initial evaluation score. Subsequently, we refine the scores of prompts producing identical answers to be mutually consistent. Experimental results show that GLaPE provides reliable evaluations uniform with accuracy, even in the absence of gold labels. Moreover, on six popular reasoning tasks, our GLaPE-based prompt optimization yields effective prompts comparable to accuracy-based ones.

  @article{zhang2024glape,
    title={GLaPE: Gold Label-agnostic Prompt Evaluation and Optimization for Large Language Model},
    author={Zhang, Xuanchang and Zhang, Zhuosheng and Zhao, Hai},
    journal={arXiv preprint arXiv:2402.02408},
    year={2024}
  }
  

• R-Judge: Benchmarking Safety Risk Awareness for LLM Agents
  Tongxin Yuan#, Zhiwei He#, Lingzhong Dong, Yiming Wang, Ruijie Zhao, Tian Xia, Lizhen Xu, Binglin Zhou, Fangqi Li, Zhuosheng Zhang*, Rui Wang, Gongshen Liu.
  EMNLP-Findings, 2024
  "Are LLM agents aware of safety risks in real-world applications? Let's find out with R-Judge!"
  [PDF] [Abstract] [Bib]
  
  R-Judge

  Large language models (LLMs) have exhibited great potential in autonomously completing tasks across real-world applications. Despite this, these LLM agents introduce unexpected safety risks when operating in interactive environments. Instead of centering on LLM-generated content safety in most prior studies, this work addresses the imperative need for benchmarking the behavioral safety of LLM agents within diverse environments. We introduce R-Judge, a benchmark crafted to evaluate the proficiency of LLMs in judging safety risks given agent interaction records. R-Judge comprises 162 agent interaction records, encompassing 27 key risk scenarios among 7 application categories and 10 risk types. It incorporates human consensus on safety with annotated safety risk labels and high-quality risk descriptions. Utilizing R-Judge, we conduct a comprehensive evaluation of 8 prominent LLMs commonly employed as the backbone for agents. The best-performing model, GPT-4, achieves 72.29% in contrast to the human score of 89.38%, showing considerable room for enhancing the risk awareness of LLMs. Notably, leveraging risk descriptions as environment feedback significantly improves model performance, revealing the importance of salient safety risk feedback. Furthermore, we design an effective chain of safety analysis technique to help the judgment of safety risks and conduct an in-depth case study to facilitate future research. R-Judge is publicly available at https://github.com/Lordog/R-Judge.

  @article{yuan2024r,
    title={R-Judge: Benchmarking Safety Risk Awareness for LLM Agents},
    author={Yuan, Tongxin and He, Zhiwei and Dong, Lingzhong and Wang, Yiming and Zhao, Ruijie and Xia, Tian and Xu, Lizhen and Zhou, Binglin and Li, Fangqi and Zhang, Zhuosheng and Wang, Rui and Liu, Gongshen},
    journal={arXiv preprint arXiv:2401.10019},
    year={2024}
  }
                

• Dynamic Planning for LLM-based Graphical User Interface Automation
  Shaoqing Zhang, Zhuosheng Zhang, Kehai Chen, Xinbe Ma, Muyun Yang, Tiejun Zhao, Min Zhang.
  EMNLP-Findings, 2024
  [PDF] [Abstract] [Bib]
  
  D-PoT

  The advent of large language models (LLMs) has spurred considerable interest in advancing autonomous LLMs-based agents, particularly in intriguing applications within smartphone graphical user interfaces (GUIs). When presented with a task goal, these agents typically emulate human actions within a GUI environment until the task is completed. However, a key challenge lies in devising effective plans to guide action prediction in GUI tasks, though planning have been widely recognized as effective for decomposing complex tasks into a series of steps. Specifically, given the dynamic nature of environmental GUIs following action execution, it is crucial to dynamically adapt plans based on environmental feedback and action this http URL show that the widely-used ReAct approach fails due to the excessively long historical dialogues. To address this challenge, we propose a novel approach called Dynamic Planning of Thoughts (D-PoT) for LLM-based GUI agents.D-PoT involves the dynamic adjustment of planning based on the environmental feedback and execution history. Experimental results reveal that the proposed D-PoT significantly surpassed the strong GPT-4V baseline by +12.7% (34.66% → 47.36%) in accuracy. The analysis highlights the generality of dynamic planning in different backbone LLMs, as well as the benefits in mitigating hallucinations and adapting to unseen tasks. Code is available at https://github.com/sqzhang-lazy/D-PoT.

  @article{zhang2024dynamic,
    title={Dynamic Planning for LLM-based Graphical User Interface Automation},
    author={Zhang, Shaoqing and Zhang, Zhuosheng and Chen, Kehai and Ma, Xinbe and Yang, Muyun and Zhao, Tiejun and Zhang, Min},
    journal={arXiv preprint arXiv:2410.00467},
    year={2024}
  }
                

• Multimodal Chain-of-Thought Reasoning in Language Models
  Zhuosheng Zhang, Aston Zhang, Mu Li, Hai Zhao, George Karypis, Alex Smola.
  TMLR, 2024
  "Imagine learning a textbook with no figures: Multimodal-CoT surpasses humans on ScienceQA."
  Featured in Dive into Deep Learning (Adopted at 500 universities from 70 countries)
  [Top Trending Research on paperswithcode] [Idea Inspiration] [PDF] [Abstract] [Bib]
  
  MM-CoT

  Large language models (LLMs) have shown impressive performance on complex reasoning by leveraging chain-of-thought (CoT) prompting to generate intermediate reasoning chains as the rationale to infer the answer. However, existing CoT studies are mostly isolated in the language modality with LLMs, where LLMs are hard to deploy. To elicit CoT reasoning in multimodality, a possible solution is to fine-tune small language models by fusing the vision and language features to perform CoT reasoning. The key challenge is that those language models tend to generate hallucinated reasoning chains that mislead the answer inference. To mitigate the effect of such mistakes, we propose Multimodal-CoT that incorporates vision features in a decoupled training framework. The framework separates the rationale generation and answer inference into two stages. By incorporating the vision features in both stages, the model is able to generate effective rationales that contribute to answer inference. With Multimodal-CoT, our model under 1 billion parameters outperforms the previous state-of-the-art LLM (GPT-3.5) by 16% (75.17%->91.68%) on the ScienceQA benchmark and even surpasses human performance. Code is publicly available at https://github.com/amazon-science/mm-cot.

  @article{zhang2023multicot,
    title={Multimodal Chain-of-Thought Reasoning in Language Models},
    author={Zhang, Zhuosheng and Zhang, Aston and Li, Mu and Zhao, Hai and Karypis, George and Smola, Alex},
    journal={arXiv preprint arXiv:2302.00923},
    year={2023}
  }
                

• Investigating Multi-Hop Factual Shortcuts in Knowledge Editing of Large Language Models
  Tianjie Ju, Yijin Chen, Xinwei Yuan, Zhuosheng Zhang*, Wei Du, Yubin Zheng, Gongshen Liu*.
  ACL, 2024
  [PDF] [Abstract] [Bib]
  
  MultiHopShortcuts

  Recent work has showcased the powerful capability of large language models (LLMs) in recalling knowledge and reasoning. However, the reliability of LLMs in combining these two capabilities into reasoning through multi-hop facts has not been widely explored. This paper systematically investigates the possibilities for LLMs to utilize shortcuts based on direct connections between the initial and terminal entities of multi-hop knowledge. We first explore the existence of factual shortcuts through Knowledge Neurons, revealing that: (i) the strength of factual shortcuts is highly correlated with the frequency of co-occurrence of initial and terminal entities in the pre-training corpora; (ii) few-shot prompting leverage more shortcuts in answering multi-hop questions compared to chain-of-thought prompting. Then, we analyze the risks posed by factual shortcuts from the perspective of multi-hop knowledge editing. Analysis shows that approximately 20% of the failures are attributed to shortcuts, and the initial and terminal entities in these failure instances usually have higher co-occurrences in the pre-training corpus. Finally, we propose erasing shortcut neurons to mitigate the associated risks and find that this approach significantly reduces failures in multiple-hop knowledge editing caused by shortcuts.

  @article{ju2024investigating,
    title={Investigating Multi-Hop Factual Shortcuts in Knowledge Editing of Large Language Models},
    author={Ju, Tianjie and Chen, Yijin and Yuan, Xinwei and Zhang, Zhuosheng and Du, Wei and Zheng, Yubin and Liu, Gongshen},
    journal={arXiv preprint arXiv:2402.11900},
    year={2024}
  }
                

• Acquiring Clean Language Models from Backdoor Poisoned Datasets by Downscaling Frequency Space
  Zongru Wu, Zhuosheng Zhang*, Pengzhou Cheng, Gongshen Liu*.
  ACL, 2024
  [PDF] [Abstract] [Bib]
  
  MuScleLoRA

  Despite the notable success of language models (LMs) in various natural language processing (NLP) tasks, the reliability of LMs is susceptible to backdoor attacks. Prior research attempts to mitigate backdoor learning while training the LMs on the poisoned dataset, yet struggles against complex backdoor attacks in real-world scenarios. In this paper, we investigate the learning mechanisms of backdoor LMs in the frequency space by Fourier analysis. Our findings indicate that the backdoor mapping presented on the poisoned datasets exhibits a more discernible inclination towards lower frequency compared to clean mapping, resulting in the faster convergence of backdoor mapping. To alleviate this dilemma, we propose Multi-Scale Low-Rank Adaptation (MuScleLoRA), which deploys multiple radial scalings in the frequency space with low-rank adaptation to the target model and further aligns the gradients when updating parameters. Through downscaling in the frequency space, MuScleLoRA encourages the model to prioritize the learning of relatively high-frequency clean mapping, consequently mitigating backdoor learning. Experimental results demonstrate that MuScleLoRA outperforms baselines significantly. Notably, MuScleLoRA reduces the average success rate of diverse backdoor attacks to below 15\% across multiple datasets and generalizes to various backbone LMs, including BERT, RoBERTa, and Llama2.

  @article{wu2024acquiring,
    title={Acquiring Clean Language Models from Backdoor Poisoned Datasets by Downscaling Frequency Space},
    author={Wu, Zongru and Zhang, Zhuosheng and Cheng, Pengzhou and Liu, Gongshen},
    journal={arXiv preprint arXiv:2402.12026},
    year={2024}
  }
  

• On the Cross-lingual Consistency of Text Watermark for Large Language Models
  Zhiwei He, Binglin Zhou, Hongkun Hao, Aiwei Liu, Xing Wang, Zhaopeng Tu, Zhuosheng Zhang, Rui Wang.
  ACL, 2024
  [PDF] [Abstract] [Bib]
  
  X-SIR

  Text watermarking technology aims to tag and identify content produced by large language models (LLMs) to prevent misuse. In this study, we introduce the concept of ''cross-lingual consistency'' in text watermarking, which assesses the ability of text watermarks to maintain their effectiveness after being translated into other languages. Preliminary empirical results from two LLMs and three watermarking methods reveal that current text watermarking technologies lack consistency when texts are translated into various languages. Based on this observation, we propose a Cross-lingual Watermark Removal Attack (CWRA) to bypass watermarking by first obtaining a response from an LLM in a pivot language, which is then translated into the target language. CWRA can effectively remove watermarks by reducing the Area Under the Curve (AUC) from 0.95 to 0.67 without performance loss. Furthermore, we analyze two key factors that contribute to the cross-lingual consistency in text watermarking and propose a defense method that increases the AUC from 0.67 to 0.88 under CWRA.

  @article{he2024can,
    title={Can Watermarks Survive Translation? On the Cross-lingual Consistency of Text Watermark for Large Language Models},
    author={He, Zhiwei and Zhou, Binglin and Hao, Hongkun and Liu, Aiwei and Wang, Xing and Tu, Zhaopeng and Zhang, Zhuosheng and Wang, Rui},
    journal={arXiv preprint arXiv:2402.14007},
    year={2024}
  }
                

• You Only Look at Screens: Multimodal Chain-of-Action Agents
  Zhuosheng Zhang, Aston Zhang.
  ACL-Findings, 2024
  "Perform a task on smart phones? Train an agent using screenshots."
  [PDF] [Abstract] [Bib] [slides]
  
  Auto-GUI

  Autonomous graphical user interface (GUI) agents aim to facilitate task automation by interacting with the user interface without manual intervention. Recent studies have investigated eliciting the capabilities of large language models (LLMs) for effective engagement in diverse environments. To align with the input-output requirement of LLMs, most existing approaches are developed under a sandbox setting where they rely on external tools and application-specific APIs to parse the environment into textual elements and interpret the predicted actions. Consequently, those approaches often grapple with inference inefficiency and error propagation risks. To mitigate the challenges, we introduce Auto-GUI, a multimodal solution that directly interacts with the interface, bypassing the need for environment parsing or reliance on application-dependent APIs. Moreover, we propose a chain-of-action technique -- leveraging a series of intermediate previous action histories and future action plans -- to help the agent decide what action to execute. We evaluate our approach on a new device-control benchmark AITW with 30K unique instructions, spanning multi-step tasks such as application operation, web searching, and web shopping. Experimental results show that Auto-GUI achieves state-of-the-art performance with an action type prediction accuracy of 90\% and an overall action success rate of 74\%. Code is publicly available at https://github.com/cooelf/Auto-GUI.

  @article{zhang2023autoui,
    title={You Only Look at Screens: Multimodal Chain-of-Action Agents},
    author={Zhang, Zhuosheng and Zhang, Aston},
    journal={arXiv preprint arXiv:2309.11436},
    year={2023}
  }
                

• Measuring Bargaining Abilities of LLMs: A Benchmark and A Buyer-Enhancement Method
  Tian Xia, Zhiwei He, Tong Ren, Yibo Miao, Zhuosheng Zhang, Yang Yang, Rui Wang.
  ACL-Findings, 2024
  [PDF] [Abstract] [Bib]
  
  AmazonPriceHistory

  Bargaining is an important and unique part of negotiation between humans. As LLM-driven agents learn to negotiate and act like real humans, how to evaluate agents' bargaining abilities remains an open problem. For the first time, we formally described the Bargaining task as an asymmetric incomplete information game, defining the gains of the Buyer and Seller in multiple bargaining processes. It allows us to quantitatively assess an agent's performance in the Bargain task. We collected a real product price dataset, AmazonHistoryPrice, and conducted evaluations of various LLM agents' bargaining abilities. We find that playing a Buyer is much harder than a Seller, and increasing model size can not effectively improve the Buyer's performance. To address the challenge, we propose a novel approach called OG-Narrator that integrates a deterministic Offer Generator to control the price range of Buyer's offers, and an LLM Narrator to create natural language sentences for generated offers. Experimental results show that OG-Narrator improves the buyer's deal rates from 26.67% to 88.88% and brings a ten times of multiplication of profits on all baselines, even a model that has not been aligned.

  @article{xia2024measuring,
    title={Measuring Bargaining Abilities of LLMs: A Benchmark and A Buyer-Enhancement Method},
    author={Xia, Tian and He, Zhiwei and Ren, Tong and Miao, Yibo and Zhang, Zhuosheng and Yang, Yang and Wang, Rui},
    journal={arXiv preprint arXiv:2402.15813},
    year={2024}
  }
                

• Comprehensive Cognitive LLM Agent for Smartphone GUI Automation
  Xinbei Ma, Zhuosheng Zhang*, Hai Zhao*.
  ACL-Findings, 2024
  [PDF] [Abstract] [Bib]
  
  CoCo-Agent

  Large language models (LLMs) have played a pivotal role in building communicative AI to imitate human behaviors but face the challenge of efficient customization. To tackle this challenge, recent studies have delved into the realm of model editing, which manipulates specific memories of language models and changes the related language generation. However, the robustness of model editing remains an open question. This work seeks to understand the strengths and limitations of editing methods, thus facilitating robust, realistic applications of communicative AI. Concretely, we conduct extensive analysis to address the three key research questions. Q1: Can edited LLMs behave consistently resembling communicative AI in realistic situations? Q2: To what extent does the rephrasing of prompts lead LLMs to deviate from the edited knowledge memory? Q3: Which knowledge features are correlated with the performance and robustness of editing? Our experimental results uncover a substantial disparity between existing editing methods and the practical application of LLMs. On rephrased prompts that are complex and flexible but common in realistic applications, the performance of editing experiences a significant decline. Further analysis shows that more popular knowledge is memorized better, easier to recall, and more challenging to edit effectively.

  @article{ma2024comprehensive,
    title={Comprehensive Cognitive LLM Agent for Smartphone GUI Automation},
    author={Ma, Xinbei and Zhang, Zhuosheng and Zhao, Hai},
    journal={arXiv preprint arXiv:2402.11941},
    year={2024}
  }
                

• Meta-Reasoning: Semantics-Symbol Deconstruction for Large Language Models
  Yiming Wang, Zhuosheng Zhang, Rui Wang.
  ACL-Findings, 2024
  [PDF] [Abstract] [Bib]
  
  Meta-Reasoning

  Neural-symbolic methods have shown their effectiveness in enhancing the reasoning abilities of large language models (LLMs). However, existing methods primarily rely on mapping natural languages to more syntactically complete formal languages (e.g., Python and SQL). Those approaches necessitate that reasoning tasks be convertible into programs, which cater more to the computer execution mindset and deviate from human reasoning habits. To expand the real-world applicability and flexibility of symbolic methods, we propose Meta-Reasoning from the scope of linguistics itself. This method empowers LLMs to deconstruct questions and effectively capture more generalized knowledge autonomously. We find that Meta-Reasoning achieves improved in-context learning efficiency, reasoning accuracy, and output stability in six arithmetic and symbolic reasoning tasks. In particular, when applied to symbolic reasoning tasks such as Tracking Shuffled Objects, GPT-3 (text-davinci-002) surpasses the few-shot Chain-of-Thought prompting approach (+37.7%), with 99% accuracy after a single demonstration of Meta-Reasoning.

  @article{wang2023meta,
    title={Meta-reasoning: Semantics-symbol deconstruction for large language models},
    author={Wang, Yiming and Zhang, Zhuosheng and Wang, Rui},
    journal={arXiv preprint arXiv:2306.17820},
    year={2023}
  }
                

• MedAgents: Large Language Models as Collaborators for Zero-shot Medical Reasoning
  Xiangru Tang#, Anni Zou#, Zhuosheng Zhang, Yilun Zhao, Xingyao Zhang, Arman Cohan, Mark Gerstein.
  ACL-Findings, 2024
  [PDF] [Abstract] [Bib]
  MedAgents

  Large Language Models (LLMs), despite their remarkable progress across various general domains, encounter significant barriers in medicine and healthcare. This field faces unique challenges such as domain-specific terminologies and the reasoning over specialized knowledge. To address these obstinate issues, we propose a novel Multi-disciplinary Collaboration (MC) framework for the medical domain that leverages role-playing LLM-based agents who participate in a collaborative multi-round discussion, thereby enhancing LLM proficiency and reasoning capabilities. This training-free and interpretable framework encompasses five critical steps: gathering domain experts, proposing individual analyses, summarising these analyses into a report, iterating over discussions until a consensus is reached, and ultimately making a decision. Our work particularly focuses on the zero-shot scenario, our results on nine data sets (MedQA, MedMCQA, PubMedQA, and six subtasks from MMLU) establish that our proposed MC framework excels at mining and harnessing the medical expertise in LLMs, as well as extending its reasoning abilities. Based on these outcomes, we further conduct a human evaluation to pinpoint and categorize common errors within our method, as well as ablation studies aimed at understanding the impact of various factors on overall performance.

  @article{tang2023medagents,
    title={MedAgents: Large Language Models as Collaborators for Zero-shot Medical Reasoning},
    author={Tang, Xiangru and Zou, Anni and Zhang, Zhuosheng and Zhao, Yilun and Zhang, Xingyao and Cohan, Arman and Gerstein, Mark},
    journal={arXiv preprint arXiv:2311.10537},
    year={2023}
  }

• Structured Chemistry Reasoning with Large Language Models
  Siru Ouyang, Zhuosheng Zhang, Bing Yan, Xuan Liu, Jiawei Han, Lianhui Qin.
  ICML, 2024
  [PDF] [Abstract] [Bib]
  
  StructChem

  This paper studies the problem of solving complex chemistry problems with large language models (LLMs). Despite the extensive general knowledge in LLMs (such as GPT-4), they struggle with chemistry reasoning that requires faithful grounded reasoning with diverse chemical knowledge and an integrative understanding of chemical interactions. We propose InstructChem, a new structured reasoning approach that substantially boosts the LLMs' chemical reasoning capabilities. InstructChem explicitly decomposes the reasoning into three critical phrases, including chemical formulae generation by LLMs that offers the basis for subsequent grounded reasoning, step-by-step reasoning that makes multi-step derivations with the identified formulae for a preliminary answer, and iterative review-and-refinement that steers LLMs to progressively revise the previous phases for increasing confidence, leading to the final high-confidence answer. We conduct extensive experiments on four different chemistry challenges, including quantum chemistry, quantum mechanics, physical chemistry, and chemistry kinetics. Our approach significantly enhances GPT-4 on chemistry reasoning, yielding an 8% average absolute improvement and a 30% peak improvement. We further use the generated reasoning by GPT-4 to fine-tune smaller LMs (e.g., Vicuna) and observe strong improvement of the smaller LMs. This validates our approach and enables LLMs to generate high-quality reasoning.

  @article{ouyang2023structured,
    title={Structured Chemistry Reasoning with Large Language Models},
    author={Ouyang, Siru and Zhang, Zhuosheng and Yan, Bing and Liu, Xuan and Han, Jiawei and Qin, Lianhui},
    journal={arXiv preprint arXiv:2311.09656},
    year={2023}
  }
  

• Self-Prompting Large Language Models for Open-Domain QA
  Junlong Li, Jinyuan Wang, Zhuosheng Zhang*, Hai Zhao*.
  NAACL, 2024
  "Free from training data and external knowledge corpus for ODQA."
  [PDF] [Abstract] [Bib]
  
  Self-Prompting

  Open-Domain Question Answering (ODQA) aims to answer questions without explicitly providing specific background documents. This task becomes notably challenging in a zero-shot setting where no data is available to train tailored retrieval-reader models. While recent Large Language Models (LLMs) like GPT-3 have demonstrated their effectiveness in zero-shot ODQA using direct prompting methods, these methods still fall short of fully harnessing the potential of LLMs when implicitly invoked. In this paper, we propose a Self-Prompting framework to explicitly utilize the massive knowledge encoded in the parameters of LLMs and their strong instruction understanding abilities. Concretely, we prompt LLMs step by step to generate multiple pseudo QA pairs with background passages and explanations entirely from scratch. These generated elements are then utilized for in-context learning. Experimental results show that our method significantly surpasses previous state-of-the-art zero-shot methods on three widely-used ODQA datasets and even achieves comparable performance with various customized fine-tuned models on full training data. Our code is available at https://github.com/lockon-n/self-prompting.

  @article{li2022self,
    title={Self-Prompting Large Language Models for Open-Domain QA},
    author={Li, Junlong and hang, Zhuosheng and Zhao, Hai},
    journal={arXiv preprint arXiv:2212.08635},
    year={2022}
  }
                

• Improving Machine Translation with Human Feedback: An Exploration of Quality Estimation as a Reward Model
  Zhiwei He, Xing Wang, Wenxiang Jiao, Zhuosheng Zhang, Rui Wang, Shuming Shi, Zhaopeng Tu.
  NAACL, 2024
  [PDF] [Abstract] [Bib]
  
  FeedbackMT

  Insufficient modeling of human preferences within the reward model is a major obstacle for leveraging human feedback to improve translation quality. Fortunately, quality estimation (QE), which predicts the quality of a given translation without reference, has achieved impressive alignment with human evaluations in the last two years. In this work, we investigate the potential of employing the QE model as the reward model (the QE-based reward model) to predict human preferences for feedback training. We first identify the overoptimization problem during QE-based feedback training, manifested as an increase in reward while translation quality declines. We examine the problem and argue that the vulnerability of the QE model might lead to high rewards for incorrect translations, resulting in overoptimization and error propagation. To address the problem, we adopt a simple yet effective method that uses heuristic rules to detect the incorrect translations and assigns a penalty term to the QE-based rewards for the detected incorrect translations. Experimental results show that the proposed QE-based feedback training achieves consistent and significant improvements across various settings, further verified through human preference studies. Our subsequent analysis demonstrates the high data efficiency of the proposed QE-based feedback training: the proposed approach using a small amount of monolingual data can outperform systems using larger parallel corpora.

  @article{he2024improving,
    title={Improving machine translation with human feedback: An exploration of quality estimation as a reward model},
    author={He, Zhiwei and Wang, Xing and Jiao, Wenxiang and Zhang, Zhuosheng and Wang, Rui and Shi, Shuming and Tu, Zhaopeng},
    journal={arXiv preprint arXiv:2401.12873},
    year={2024}
  }
                

• Mitigating Harmful Chain-of-Thought Reasoning with Selective Filtering
  Yexin Wu, Zhuosheng Zhang*, Hai Zhao*.
  LREC-COLING, 2024
  [PDF] [Abstract] [Bib]
  
  SelF-Reasoner

  Large language models have manifested remarkable capabilities by leveraging chain-of-thought (CoT) reasoning techniques to solve intricate questions through step-by-step reasoning chains. Despite its success, the efficacy of such reasoning is inherently contingent upon the quality of CoT. However, flawless CoT reasoning cannot be guaranteed due to the presence of indecomposable questions and the potential for erroneous reasoning chains, particularly in the case of small-scale language models. To tackle this challenge, we propose a novel approach called the selective filtering reasoner (SelF-Reasoner) that assesses the entailment relationship between the question and the candidate reasoning chain. Then, we proceed with CoT reasoning when the reasoning chain demonstrates confidence; otherwise, we opt to predict the answer directly. SelF-Reasoner improves the fine-tuned T5 baseline consistently over the ScienceQA, ECQA, and LastLetter tasks. Code is available at https://github.com/LibroWu/SelF-Reasoner.

                  @article{wu2024mitigating,
                    title={Mitigating Misleading Chain-of-Thought Reasoning with Selective Filtering},
                    author={Wu, Yexin and Zhang, Zhuosheng and Zhao, Hai},
                    journal={arXiv preprint arXiv:2403.19167},
                    year={2024}
                  }
                

• AuRoRA: A One-for-all Platform for Augmented Reasoning and Refining with Task-Adaptive Chain-of-Thought Prompting
  Anni Zou, Zhuosheng Zhang*, Hai Zhao*.
  LREC-COLING, 2024
  [PDF] [Abstract] [Bib] [Demo]
  

  Large language models (LLMs) empowered by chain-of-thought (CoT) prompting have yielded remarkable prowess in reasoning tasks. Nevertheless, current methods predominantly lean on handcrafted or task-specific demonstrations, lack reliable knowledge basis and thus struggle for trustworthy responses in an automated pattern. While recent works endeavor to improve upon one certain aspect, they ignore the importance and necessity of establishing an integrated and interpretable reasoning system. To address these drawbacks and provide a universal solution, we propose \textsc{AuRoRA}: a one-for-all platform for augmented reasoning and refining based on CoT prompting that excels in adaptability, reliability, integrity, and interpretability. The system exhibits superior performances across six reasoning tasks and offers real-time visual analysis, which has pivotal academic and application value in the era of LLMs.

• Multi-turn Dialogue Comprehension from a Topic-aware Perspective
  Xinbei Ma, Yi Xu, Hai Zhao, Zhuosheng Zhang.
  Neurocomputing, 2024
  [PDF] [Abstract] [Bib]
  

  Dialogue related Machine Reading Comprehension requires language models to effectively decouple and model multi-turn dialogue passages. As a dialogue development goes after the intentions of participants, its topic may not keep constant through the whole passage. Hence, it is non-trivial to detect and leverage the topic shift in dialogue modeling. Topic modeling, although has been widely studied in plain text, deserves far more utilization in dialogue reading comprehension. This paper proposes to model multi-turn dialogues from a topic-aware perspective. We start with a dialogue segmentation algorithm to split a dialogue passage into topic-concentrated fragments in an unsupervised way. Then we use these fragments as topic-aware language processing units in further dialogue comprehension. On one hand, the split segments indict specific topics rather than mixed intentions, thus showing convenient on in-domain topic detection and location. For this task, we design a clustering system with a self-training auto-encoder, and we build two constructed datasets for evaluation. On the other hand, the split segments are an appropriate element of multi-turn dialogue response selection. For this purpose, we further present a novel model, Topic-Aware Dual-Attention Matching (TADAM) Network, which takes topic segments as processing elements and matches response candidates with a dual cross-attention. Empirical studies on three public benchmarks show great improvements over baselines. Our work continues the previous studies on document topic, and brings the dialogue modeling to a novel topic-aware perspective with exhaustive experiments and analyses.

  @article{ma2023multi,
    title={Multi-turn Dialogue Comprehension from a Topic-aware Perspective},
    author={Ma, Xinbei and Xu, Yi and Zhao, Hai and Zhang, Zhuosheng},
    journal={arXiv preprint arXiv:2309.09666},
    year={2023}
  }
                

• Fact-driven Logical Reasoning for Machine Reading Comprehension
  Siru Ouyang, Zhuosheng Zhang*, Hai Zhao*.
  AAAI, 2024
  [PDF] [Abstract] [Bib]
  
  FocalReasoner

  Recent years have witnessed an increasing interest in training machines with reasoning ability, which deeply relies on accurately and clearly presented clue forms. The clues are usually modeled as entity-aware knowledge in existing studies. However, those entity-aware clues are primarily focused on commonsense, making them insufficient for tasks that require knowledge of temporary facts or events, particularly in logical reasoning for reading comprehension. To address this challenge, we are motivated to cover both commonsense and temporary knowledge clues hierarchically. Specifically, we propose a general formalism of knowledge units by extracting backbone constituents of the sentence, such as the subject-verb-object formed ``facts''. We then construct a supergraph on top of the fact units, allowing for the benefit of sentence-level (relations among fact groups) and entity-level interactions (concepts or actions inside a fact). Experimental results on logical reasoning benchmarks and dialogue modeling datasets show that our approach improves the baselines substantially, and it is general across backbone models.

  @article{ouyang2024fact,
    title={Fact-driven Logical Reasoning for Machine Reading Comprehension},
    author={Ouyang, Siru and Zhang, Zhuosheng and Zhao, Hai},
    journal={The 38th Annual AAAI Conference on Artificial Intelligence (AAAI 2024)},
    year={2024}
  }
                

• Exploring Human-Like Translation Strategy with Large Language Models
  Zhiwei He, Tian Liang, Wenxiang Jiao, Zhuosheng Zhang, Yujiu Yang, Rui Wang, Zhaopeng Tu, Shuming Shi, Xing Wang.
  TACL, 2023
  Delves into LLMs' potential for mimicking human translation strategies.
  [PDF] [Abstract] [Bib]
  
  MAPS-mt

  Large language models (LLMs) have demonstrated impressive capabilities in general scenarios, exhibiting a level of aptitude that approaches, in some aspects even surpasses, human-level intelligence. Among their numerous skills, the translation abilities of LLMs have received considerable attention. In contrast to traditional machine translation that focuses solely on source-target mapping, LLM-based translation can potentially mimic the human translation process that takes many preparatory steps to ensure high-quality translation. This work aims to explore this possibility by proposing the MAPS framework, which stands for Multi-Aspect Prompting and Selection. Specifically, we enable LLMs to first analyze the given source text and extract three aspects of translation-related knowledge: keywords, topics and relevant demonstrations to guide the translation process. To filter out the noisy and unhelpful knowledge, we employ a selection mechanism based on quality estimation. Experiments suggest that MAPS brings significant and consistent improvements over text-davinci-003 and Alpaca on eight translation directions from the latest WMT22 test sets. Our further analysis shows that the extracted knowledge is critical in resolving up to 59% of hallucination mistakes in translation. Code is available at this https https://github.com/zwhe99/MAPS-mt.

  @article{he2023exploring,
    title={Exploring Human-Like Translation Strategy with Large Language Models},
    author={He, Zhiwei and Liang, Tian and Jiao, Wenxiang and Zhang, Zhuosheng and Yang, Yujiu and Wang, Rui and Tu, Zhaopeng and Shi, Shuming and Wang, Xing},
    journal={arXiv preprint arXiv:2305.04118},
    year={2023}
  }
                

• Is ChatGPT a General-Purpose Natural Language Processing Task Solver?
  Chengwei Qin, Aston Zhang, Zhuosheng Zhang, Jiaao Chen, Michihiro Yasunaga, Diyi Yang
  EMNLP, 2023
  "Benchmarking ChatGPT on 20 popular NLP datasets covering 7 representative task categories."
  [PDF] [Abstract] [Bib]
  

  Spurred by advancements in scale, large language models (LLMs) have demonstrated the ability to perform a variety of natural language processing (NLP) tasks zero-shot -- i.e., without adaptation on downstream data. Recently, the debut of ChatGPT has drawn a great deal of attention from the natural language processing (NLP) community due to the fact that it can generate high-quality responses to human input and self-correct previous mistakes based on subsequent conversations. However, it is not yet known whether ChatGPT can serve as a generalist model that can perform many NLP tasks zero-shot. In this work, we empirically analyze the zero-shot learning ability of ChatGPT by evaluating it on 20 popular NLP datasets covering 7 representative task categories. With extensive empirical studies, we demonstrate both the effectiveness and limitations of the current version of ChatGPT. We find that ChatGPT performs well on many tasks favoring reasoning capabilities (e.g., arithmetic reasoning) while it still faces challenges when solving specific tasks such as sequence tagging. We additionally provide in-depth analysis through qualitative case studies.

  @article{qin2023chatgpt,
    title={Is ChatGPT a General-Purpose Natural Language Processing Task Solver?},
    author={Qin, Chengwei and Zhang, Aston and Zhang, Zhuosheng and Chen, Jiaao and Yasunaga, Michihiro and Yang, Diyi},
    journal={arXiv preprint arXiv:2302.06476},
    year={2023}
  }
                

• Learning Better Masking for Better Language Model Pre-training
  Dongjie Yang, Zhuosheng Zhang*, Hai Zhao*.
  ACL, 2023
  [PDF] [Abstract] [Bib]
  
  BetterMasking

  Masked Language Modeling (MLM) has been widely used as the denoising objective in pre-training language models (PrLMs). Existing PrLMs commonly adopt a Random-Token Masking strategy where a fixed masking ratio is applied and different contents are masked by an equal probability throughout the entire training. However, the model may receive complicated impact from pre-training status, which changes accordingly as training time goes on. In this paper, we show that such time-invariant MLM settings on masking ratio and masked content are unlikely to deliver an optimal outcome, which motivates us to explore the influence of time-variant MLM settings. We propose two scheduled masking approaches that adaptively tune the masking ratio and masked content in different training stages, which improves the pre-training efficiency and effectiveness verified on the downstream tasks. Our work is a pioneer study on time-variant masking strategy on ratio and content and gives a better understanding of how masking ratio and masked content influence the MLM pre-training.

  @inproceedings{yang2023learning,
    title={Learning Better Masking for Better Language Model Pre-training},
    author={Yang, Dongjie and Zhang, Zhuosheng and Zhao, Hai},
    booktitle={The 61st Annual Meeting of the Association for Computational Linguistics (ACL 2023)},
    year={2023}
  }
                

• On Element-aware Automatic Summarization: Expert-writing Test Set and Chain-of-Thought Method
  Yiming Wang, Zhuosheng Zhang, Rui Wang.
  ACL, 2023
  "You really need higher-quality reference summaries to evaluate LLMs!"
  [PDF] [Abstract] [Bib]
  
  SumCoT

  Automatic summarization generates concise summaries that contain key ideas of source documents. As the most mainstream datasets for the news sub-domain, CNN/DailyMail and BBC XSum have been widely used for performance benchmarking. However, the reference summaries of those datasets turn out to be noisy, mainly in terms of factual hallucination and information redundancy. To address this challenge, we first annotate new expert-writing Element-aware test sets following the "Lasswell Communication Model" proposed by Lasswell (1948), allowing reference summaries to focus on more fine-grained news elements objectively and comprehensively. Utilizing the new test sets, we observe the surprising zero-shot summary ability of LLMs, which addresses the issue of the inconsistent results between human preference and automatic evaluation metrics of LLMs' zero-shot summaries in prior work. Further, we propose a Summary Chain-of-Thought (SumCoT) technique to elicit LLMs to generate summaries step by step, which helps them integrate more fine-grained details of source documents into the final summaries that correlate with the human writing mindset. Experimental results show our method outperforms state-of-the-art fine-tuned PLMs and zero-shot LLMs by +4.33/+4.77 in ROUGE-L on the two datasets, respectively. Dataset and code are publicly available at this https https://github.com/Alsace08/SumCoT.

  @inproceedings{wang2023element,
    title={Element-aware Summarization with Large Language Models: Expert-aligned Evaluation and Chain-of-Thought Method},
    author={Wang, Yiming and Zhang, Zhuosheng and Wang, Rui},
    booktitle={The 61st Annual Meeting of the Association for Computational Linguistics (ACL 2023)},
    year={2023}
  }
                

• Decker: Double Check with Heterogeneous Knowledge for Commonsense Fact Verification
  Anni Zou, Zhuosheng Zhang, Hai Zhao.
  ACL-Findings, 2023
  [PDF] [Abstract] [Bib]
  
  Decker

  Commonsense fact verification, as a challenging branch of commonsense question-answering (QA), aims to verify through facts whether a given commonsense claim is correct or not. Answering commonsense questions necessitates a combination of knowledge from various levels. However, existing studies primarily rest on grasping either unstructured evidence or potential reasoning paths from structured knowledge bases, yet failing to exploit the benefits of heterogeneous knowledge simultaneously. In light of this, we propose Decker, a commonsense fact verification model that is capable of bridging heterogeneous knowledge by uncovering latent relationships between structured and unstructured knowledge. Experimental results on two commonsense fact verification benchmark datasets, CSQA2.0 and CREAK demonstrate the effectiveness of our Decker and further analysis verifies its capability to seize more precious information through reasoning.

  @inproceedings{zou2023decker,
    title={Decker: Double Check with Heterogeneous Knowledge for Commonsense Fact Verification},
    author={Zou, Anni and Zhang, Zhuosheng and Zhao, Hai},
    booktitle={The 61st Annual Meeting of the Association for Computational Linguistics (ACL 2023)},
    year={2023}
  }
                

• Automatic Chain of Thought Prompting in Large Language Models
  Zhuosheng Zhang, Aston Zhang, Mu Li, Alex Smola.
  ICLR, 2023
  "Let's think not just step by step, but also one by one."
  Featured in Dive into Deep Learning (Adopted at 400 universities from 60 countries)
  [PDF] [Abstract] [Bib] [bilibili] [slides]
  
  Auto-CoT

  Large language models (LLMs) can perform complex reasoning by generating intermediate reasoning steps. Providing these steps for prompting demonstrations is called chain-of-thought (CoT) prompting. CoT prompting has two major paradigms. One leverages a simple prompt like "Let's think step by step" to facilitate step-by-step thinking before answering a question. The other uses a few manual demonstrations one by one, each composed of a question and a reasoning chain that leads to an answer. The superior performance of the second paradigm hinges on the hand-crafting of task-specific demonstrations one by one. We show that such manual efforts may be eliminated by leveraging LLMs with the "Let's think step by step" prompt to generate reasoning chains for demonstrations one by one, i.e., let's think not just step by step, but also one by one. However, these generated chains often come with mistakes. To mitigate the effect of such mistakes, we find that diversity matters for automatically constructing demonstrations. We propose an automatic CoT prompting method: Auto-CoT. It samples questions with diversity and generates reasoning chains to construct demonstrations. On ten public benchmark reasoning tasks with GPT-3, Auto-CoT consistently matches or exceeds the performance of the CoT paradigm that requires manual designs of demonstrations. Code is available at https://github.com/amazon-research/auto-cot

  @inproceedings{zhang2023automatic,
    title={Automatic Chain of Thought Prompting in Large Language Models},
    author={Zhang, Zhuosheng and Zhang, Aston and Li, Mu and Smola, Alex},
    booktitle={The Eleventh International Conference on Learning Representations (ICLR 2023)},
    year={2023}
  }
                

• Enhanced Speaker-aware Multi-party Multi-turn Dialogue Comprehension
  Xinbei Ma, Zhuosheng Zhang*, Hai Zhao*.
  TASLP, 2023
  [PDF] [Abstract] [Bib]
  
  ESA

  Multi-party multi-turn dialogue comprehension brings unprecedented challenges in handling complicated scenarios, as the co-occurrence of multiple speakers causes complexity and inconsistency. As a result of the multiple participation, the shift of speaker roles and crisscrossed discourse relations among utterances hinder reading comprehension. Motivated by this, we further integrate the enhancements of speaker-related features for dialogue comprehension performance. This work proposes a novel model with enhancement from both sides of speaker roles and speaker-aware relations. At the token level, we apply a speaker mask for attention, while at the discourse level, we utilize heterogeneous graph networks for comprehensive speaker-aware discourse clues. Experimental results show that our E nhanced S peaker- A ware method (ESA) helps achieve state-of-the-art performance on the Molweni dataset, as well as significant improvements on the FriendsQA dataset. We find that our method makes steady improvements on stronger backbones. Analysis shows that our model enhances the connections between utterances and their own speakers and captures the speaker-aware discourse relations. Discussions on data features and error cases are presented, and a visualized case is displayed. The findings reveal the importance of speaker-aware signals in dialogue comprehension.

  @ARTICLE{10147329,
    author={Ma, Xinbei and Zhang, Zhuosheng and Zhao, Hai},
    journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing}, 
    title={Enhanced Speaker-aware Multi-party Multi-turn Dialogue Comprehension}, 
    year={2023},
    volume={},
    number={},
    pages={1-16},
    doi={10.1109/TASLP.2023.3284516}
  }
                

• Universal Multimodal Representation for Language Understanding
  Zhuosheng Zhang#, Kehai Chen, Rui Wang#, Masao Utiyama, Eiichiro Sumita, Zuchao Li, Hai Zhao
  TPAMI, 2023
  "Let's retrieve images to overcome the lack of large-scale bilingual pairs."
  [PDF] [Abstract] [Bib]
  UVR-NMT

  Representation learning is the foundation of natural language processing (NLP). This work presents new methods to employ visual information as assistant signals to general NLP tasks. For each sentence, we first retrieve a flexible number of images either from a light topic-image lookup table extracted over the existing sentence-image pairs or a shared cross-modal embedding space that is pre-trained on out-of-shelf text-image pairs. Then, the text and images are encoded by a Transformer encoder and convolutional neural network, respectively. The two sequences of representations are further fused by an attention layer for the interaction of the two modalities. In this study, the retrieval process is controllable and flexible. The universal visual representation overcomes the lack of large-scale bilingual sentence-image pairs. Our method can be easily applied to text-only tasks without manually annotated multimodal parallel corpora. We apply the proposed method to a wide range of natural language generation and understanding tasks, including neural machine translation, natural language inference, and semantic similarity. Experimental results show that our method is generally effective for different tasks and languages. Analysis indicates that the visual signals enrich textual representations of content words, provide fine-grained grounding information about the relationship between concepts and events, and potentially conduce to disambiguation.

  @ARTICLE{zhang2023universal,
    author={Zhang, Zhuosheng and Chen, Kehai and Wang, Rui and Utiyama, Masao and Sumita, Eiichiro and Li, Zuchao and Zhao, Hai},
    journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
    title={Universal Multimodal Representation for Language Understanding}, 
    year={2023},
    volume={},
    number={},
    pages={1-18},
    doi={10.1109/TPAMI.2023.3234170}}
                

• Language Model Pre-training on True Negatives
  Zhuosheng Zhang, Hai Zhao, Masao Utiyama, Eiichiro Sumita
  AAAI, 2023
  [PDF] [Abstract] [Bib]

  Discriminative pre-trained language models (PLMs) learn to predict original texts from intentionally corrupted ones. Taking the former text as positive and the latter as negative samples, the PLM can be trained effectively for contextualized representation. However, the training of such a type of PLMs highly relies on the quality of the automatically constructed samples. Existing PLMs simply treat all corrupted texts as equal negative without any examination, which actually lets the resulting model inevitably suffer from the false negative issue where training is carried out on pseudo-negative data and leads to less efficiency and less robustness in the resulting PLMs. In this work, on the basis of defining the false negative issue in discriminative PLMs that has been ignored for a long time, we design enhanced pre-training methods to counteract false negative predictions and encourage pre-training language models on true negatives by correcting the harmful gradient updates subject to false negative predictions. Experimental results on GLUE and SQuAD benchmarks show that our counter-false-negative pre-training methods indeed bring about better performance together with stronger robustness.

  @inproceedings{zhang2023TrueNeg,
    title={Language Model Pre-training on True Negatives},
    author={Zhang, Zhuosheng and Zhao, Hai and Utiyama, Masao and Sumita, Eiichiro},
    booktitle={The Thirty-Seventh AAAI Conference on Artificial Intelligence (AAAI 2023)},
    year={2023}
  }
                

Shared Tasks

• The best among all submissions.
[Leaderboard] [Paper]

• The best among all submissions.
[Leaderboard] [Paper]

• The best among all submissions.
[Leaderboard] [Paper]

• The best models for both single and ensemble settings among all submissions (2020.01).
• The first to surpass human benchmark on both EM and F1 scores with a single model (from 2019.07-09).
• The first time to exceed 90% F1 score with ensemble models.
  [Leaderboard] [Paper] [Report]

• The best among all submissions.
• The best among all academic submissions.
  [Leaderboard] [Paper] [Report]

• The best among all submissions.
  [Leaderboard] [Paper]

• The 3rd best among all submissions.
• The best among all academic submissions.
  [Leaderboard] [Paper]

• The best single system and the second ensemble system (Silver Medal).
  [Leaderboard] [Paper] [Source]

Awards & Honors

• 2024: WAIC Youth Outstanding Paper Award, World Artificial Intelligence Conference.

• 2024: WAIC YunFan Award: Bright Star, World Artificial Intelligence Conference.

• 2023: Excellent Doctoral Thesis of Chinese Information Processing Society (CIPS).

• 2023: Shanghai Outstanding Doctoral Graduate.

• 2022: Academic Stars of Graduate Students (10 recipients), Shanghai Jiao Tong University.

• 2021: Global Top 100 Chinese Rising Stars in Artificial Intelligence (Top 10 recommended), Baidu Research.

• 2021: Baidu Scholarship (10 recipients, worldwide), Baidu.

• 2020: National Scholarship of China, Ministry of Education of the P.R. China.

• 2019: Yang Yuanqing Education Fund, The foundation of Class 1988 in CS @ Shanghai Jiao Tong University.

• 2018: Academic Stars of Graduate Students (The only master student awardee), Shanghai Jiao Tong University.

• 2016: National Figures Nomination of College Students (20 total recipients), Ministry of Education of the P.R. China.

• 2015: CCF Elite Collegiate Award, China Computer Federation.

Teaching

• NIS8021: Frontier Technology in Natural Language Processing
  Graduate, Shanghai Jiao Tong University, Fall 2024.
• NIS3353: Artificial Intelligence Security
  Undergraduate, Shanghai Jiao Tong University, Spring 2024.

Academic Service

• Organization:
  • Session Chair at RL China 2024.
  • Session Chair at CJNLP 2024.
  • Session Chair at IJCNLP-AACL 2023.
  • Co-chair of CCL Student Seminar, 2022
  • President of IBM Tech Club at Wuhan University, 2014-2015.
• Area Chair / Action Editor/ SPC:
  • ACL Rolling Review
  • LREC-COLING 2024
  • IJCAI 2024
  • ICLR 2023 TinyPapers
• Program Committee Member:
  • ML/AI conferences: ICLR, ICML, NeurIPS, AAAI, IJCAI, etc.
  • CL/NLP conferences: ARR, ACL, EMNLP, COLING, NAACL, AACL, NLPCC, CCL, etc.

• Journal Reviewer: Artificial Intelligence, IEEE/ACM TASLP, IEEE TNNLS, IEEE TETCI, IEEE Communications Magazine, ACM TALLIP, ACM TOIS, TMLR, Neurocomputing, Multimedia Systems, Neural Computing and Applications, Expert Systems With Applications.

Experience

• Jul. 2022 - Aug. 2023, Amazon Web Services AI, CA, USA.
  Applied Scientist Intern (remote), advised by Dr. Aston Zhang, Mu Li, Alex Smola.
• Feb. 2022 - June. 2022, Microsoft Cognitive Services Research Group, WA, USA.
  Research Intern (remote), advised by Dr. Shuohang Wang.
• Mar. 2021 - Dec. 2021, Langboat Tech, Beijing, China.
  Research Intern (remote), advised by Prof. Ming Zhou.
• Jun. 2019 - Jul. 2020, NICT, Kyoto, Japan.
  Internship Research Fellow, advised by Prof. Rui Wang, Kehai Chen, Masao Utiyama, and Eiichiro Sumita.

Education

• Sept. 2020 - Sept. 2023
  Ph.D., Dept. of Computer Science and Engineering, Shanghai Jiao Tong University, advised by Prof. Hai Zhao.
• Sept. 2016 - Mar. 2020
  M.S., Dept. of Computer Science and Engineering, Shanghai Jiao Tong University, advised by Prof. Hai Zhao.
• Sept. 2012 - Jun. 2016
  B.S., Dept. of Computer Science and Engineering, Wuhan University, advised by Prof. Haojun Ai.