Papers by Zheng Cao
Chinese SafetyQA: A Safety Short-form Factuality Benchmark for Large Language Models (2025.acl-long)
Copied to clipboard
Yingshui Tan, Boren Zheng, Baihui Zheng, Kerui Cao, Huiyun Jing, Jincheng Wei, Jiaheng Liu, Yancheng He, Wenbo Su, Xiaoyong Zhu, Bo Zheng, Kaifu Zhang
| Challenge: | Large language models have created significant safety concerns . factuality ability is crucial in determining whether they can be deployed and applied safely and compliantly within specific regions. |
| Approach: | They propose a benchmark to evaluate the factuality of large language models in China . they evaluate the models' ability to provide accurate and reliable information . |
| Outcome: | The proposed benchmark evaluates the factuality abilities of existing LLMs and compares them to LLM abilities. |
ScaleBox: Enabling High-Fidelity and Scalable Code Verification for Large Language Models (2026.acl-demo)
Copied to clipboard
Jiasheng Zheng, Xin Zheng, Boxi Cao, Pengbo Wang, Zhengzhao Ma, Qiming Zhu, Jiazhen Jiang, Yaojie Lu, Hongyu Lin, Xianpei Han, Le Sun
| Challenge: | Existing code sandboxes fail to provide accurate verification and efficiency under high-concurrency workloads. |
| Approach: | They propose a high-fidelity code verification system that provides sandbox feedback for RL training and evaluation. |
| Outcome: | The proposed system outperforms heuristic-matching baselines on LiveCodeBench and training stability on high-concurrency workloads. |
All Languages Matter: Understanding and Mitigating Language Bias in Multilingual RAG (2026.acl-long)
Copied to clipboard
Dan Wang, Guozhao Mo, Yafei Shi, Cheng Zhang, Bo Zheng, Boxi Cao, Xuanang Chen, Yaojie Lu, Hongyu Lin, Ben He, Xianpei Han, Le Sun
| Challenge: | Existing mRAG systems suffer from a language bias during reranking, systematically favoring English and the query’s native language. |
| Approach: | They propose a language-agnostic utility-driven reranker alignment technique to mitigate language bias during re-ranking. |
| Outcome: | The proposed approach mitigates language bias and consistently improves mRAG performance across languages. |
Unified Thinker: A General Reasoning Core for Image Generation (2026.acl-long)
Copied to clipboard
Sashuai Zhou, Qiang Zhou, Jijin Hu, Hanqing Yang, Yue Cao, Junpeng Ma, Yinchao Ma, Jun Song, Tiezheng Ge, Cheng Yu, Bo Zheng, Zhou Zhao
| Challenge: | generative models struggle with logic-intensive instruction following, exposing a persistent reasoning–execution gap. |
| Approach: | They propose a task-agnostic reasoning architecture for general image generation . they propose pixel-level feedback to ground the Thinker's policy in pixel feedback . |
| Outcome: | The proposed system significantly improves image reasoning and generation quality. |
DialogQAE: N-to-N Question Answer Pair Extraction from Customer Service Chatlog (2023.findings-emnlp)
Copied to clipboard
Xin Zheng, Tianyu Liu, Haoran Meng, Xu Wang, Yufan Jiang, Mengliang Rao, Binghuai Lin, Yunbo Cao, Zhifang Sui
| Challenge: | Existing work on question-answer extraction fails to integrate incomplete utterances from dialog context for composite QA retrieval. |
| Approach: | They propose a task where questions and corresponding answers might be separated across different utterances. |
| Outcome: | The proposed methods perform well on 5 customer service datasets and set a benchmark for N-to-N DialogQAE with utterance and session level evaluation metrics. |
Gaussian Process based Deep Dyna-Q approach for Dialogue Policy Learning (2021.findings-acl)
Copied to clipboard
| Challenge: | Reinforcement learning (RL) is the main dialogue policy learning method in recent years. |
| Approach: | They propose a Gaussian Process based Deep Dyna-Q approach to dialogue policy learning . they propose evaluating the quality of experiences generated by the world model using a discriminator . |
| Outcome: | The proposed approach improves the effectiveness and efficiency of dialogue policy learning by 20% with fewer human-machine interactions. |
MemoBrain: Executive Memory as an Agentic Brain for Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are inherently long-horizon, causing reasoning traces and tool artifacts to accumulate and strain the working context of large language models. |
| Approach: | They propose a model that constructs a dependency-aware memory over reasoning steps and captures salient intermediate states and their logical relations. |
| Outcome: | The proposed model prunes invalid steps, folds completed sub-trajectories, and preserves a compact, high-salience reasoning backbone under a fixed context budget. |
AWARE: Agentic Knowledge Warehousing for Contextual Intelligence (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models excel in information seeking tasks, but their knowledge is limited in coverage and timeliness. |
| Approach: | They propose an agentic knowledge warehousing framework that transforms unstructured data into minimal, task-conditioned knowledge representations consumable by LLMs. |
| Outcome: | Experiments on GAIA, WebWalker, and BrowseComp-Plus show improvements over baselines. |
Doc2EDAG: An End-to-End Document-level Framework for Chinese Financial Event Extraction (D19-1)
Copied to clipboard
| Challenge: | Existing event extraction methods are limited to extract event arguments within the sentence scope. |
| Approach: | They propose a model which generates an entity-based directed acyclic graph to fulfill document-level EE effectively. |
| Outcome: | The proposed model can generate entity-based directed acyclic graph to fulfill document-level EE effectively. |
RetroMAE-2: Duplex Masked Auto-Encoder For Pre-Training Retrieval-Oriented Language Models (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods for retrieval-oriented language models focus on contextualized embedding of the [CLS] token, but recent study shows that ordinary tokens besides [CLL] may provide extra information, which help to produce a better representation effect. |
| Approach: | They propose a method where all contextualized embeddings of pre-trained model can be jointly pre-trained for retrieval tasks. |
| Outcome: | The proposed method improves the quality of representation where all contextualized embeddings of the pre-trained model can be leveraged. |
CSS: A Large-scale Cross-schema Chinese Text-to-SQL Medical Dataset (2023.findings-acl)
Copied to clipboard
| Challenge: | a cross-domain text-to-SQL task aims to parse user questions into SQL on complete unseen databases . a single-domain task evaluates the performance on identical databases based on the same domain . |
| Approach: | They propose a cross-domain text-to-SQL task that parses user questions into SQL on unseen databases. |
| Outcome: | The proposed system can parse user questions into SQL on complete unseen databases. |
Exploring the Impact of Negative Samples of Contrastive Learning: A Case Study of Sentence Embedding (2022.findings-acl)
Copied to clipboard
| Challenge: | Unsupervised contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data. |
| Approach: | They propose a momentum contrastive learning model with negative sample queue for sentence embedding with a simulated model with EMA update mechanism. |
| Outcome: | The proposed model achieves a Spearman’s correlation of 77.27% on the semantic text similarity task and a maximum traceable distance metric. |
Depth Aware Hierarchical Replay Continual Learning for Knowledge Based Question Answering (2024.lrec-main)
Copied to clipboard
| Challenge: | Continual learning models adapt well to the latest data but lose ability to remember past data due to changes in the data source. |
| Approach: | They propose a hierarchical replay framework that allows models to keep a small memory of previous learned data that uses replay. |
| Outcome: | The proposed model outperforms previous continual learning methods in mitigating catastrophic forgetting. |
MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing (2026.acl-industry)
Copied to clipboard
Junbo Niu, Zheng Liu, Zhuangcheng Gu, Bin Wang, Linke Ouyang, Zhiyuan Zhao, Tao Chu, Tianyao He, Fan Wu, Qintong Zhang, Zhenjiang Jin, Guang Liang, Rui Zhang, Wenzheng Zhang, Yuan Qu, Zhifei Ren, Yuefeng Sun, Zirui Tang, Boyu Niu, Yuanhong Zheng, Dongsheng Ma, Ziyang Miao, Hejun Dong, Siyi Qian, Junyuan Zhang, Fangdong Wang, Jingzhou Chen, Xiaomeng Zhao, Liqun Wei, Wei Li, Shasha Wang, RuiLiang Xu, Yuanyuan Cao, Lu Chen, Qianqian Wu, Huaiyu Gu, Lindong Lu, Dechen Lin, null Shenguanlin, Xuanhe Zhou, Linfeng Zhang, Yuhang Zang, Xiaoyi Dong, Jiaqi Wang, Bo Zhang, Lei Bai, Pei Chu, Weijia Li, Jiang Wu, Lijun Wu, Zhenxiang Li, Guangyu Wang, Zhongying Tu, Chao Xu, Kai Chen, Bowen Zhou, Dahua Lin, Wentao Zhang, Conghui He
| Challenge: | Document images are characterized by higher resolutions, denser content, and more complex structural layouts. |
| Approach: | They propose a 1.2B-parameter document parsing vision-language model that decouples layout analysis from local content recognition. |
| Outcome: | The proposed model surpasses general-purpose and domain-specific models on multiple benchmarks while maintaining significantly lower computational overhead. |
InquireMobile: Teaching VLM-based Mobile Agent to Request Human Assistance via Reinforcement Fine-Tuning (2026.acl-long)
Copied to clipboard
Qihang Ai, Pi Bu, Yue Cao, Yingyao Wang, Jihao Gu, Jingxuan Xing, Zekun Zhu, Wei Jiang, Zhicheng Zheng, Jun Song, Yuning Jiang
| Challenge: | Recent advances in Vision-Language Models (VLMs) have enabled mobile agents to perceive and interact with real-world mobile environments based on human instructions. |
| Approach: | They propose a vision-language model that actively seeks human confirmation at critical decision points and a model inspired by reinforcement learning. |
| Outcome: | The proposed model achieves an improvement of 46.8% in inquiry success rate and the best overall success rate among existing baselines on InquireBench. |
Unlocking Memorization in Large Language Models with Dynamic Soft Prompting (2024.emnlp-main)
Copied to clipboard
Zhepeng Wang, Runxue Bao, Yawen Wu, Jackson Taylor, Cao Xiao, Feng Zheng, Weiwen Jiang, Shangqian Gao, Yanfu Zhang
| Challenge: | Pretrained large language models excel in a variety of natural language processing tasks . however, they pose significant security risks due to their tendency to memorize training data . |
| Approach: | They propose a method to estimate LLM memorization using dynamic, prefix-dependent soft prompts. |
| Outcome: | The proposed method can achieve maximum relative improvement of 135.3% and 39.8% over baseline compared to state-of-the-art methods. |
Multilingual Knowledge Graph Completion with Self-Supervised Adaptive Graph Alignment (2022.acl-long)
Copied to clipboard
Zijie Huang, Zheng Li, Haoming Jiang, Tianyu Cao, Hanqing Lu, Bing Yin, Karthik Subbian, Yizhou Sun, Wei Wang
| Challenge: | Existing methods to predict missing facts in knowledge graphs are limited in language alignment . SS-AGA uses seed alignment as an edge type to fuses all KGs as a whole graph . |
| Approach: | They propose a self-supervised adaptive graph alignment method that fuses all KGs as a whole graph by regarding alignment as 'a new edge type' they propose SS-AGA method that uses relation-aware attention weights to capture potential alignment pairs in a new paradigm. |
| Outcome: | The proposed method can predict missing facts in a knowledge graph (KG) but language alignment is scarce and new alignment identification is noisy. |
Interactive Learning for LLM Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing multi-agent learning approaches foster collaboration among Large Language Models (LLMs) yet they still rely on re-executing the MAS during inference. |
| Approach: | They propose a co-learning framework that integrates Dynamic Interaction and Perception Calibration to enhance LLMs' independent problem-solving ability. |
| Outcome: | The proposed framework integrates Dynamic Interaction and Perception Calibration to improve LLMs' independent problem-solving ability. |
PGPO: Enhancing Agent Reasoning via Pseudocode-style Planning Guided Preference Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing LLM agents generate verbose and inefficient natural language plans to guide reasoning, which restricts agents’ ability to generalize across similar tasks. |
| Approach: | They propose a pseudocode-style planning guide optimization method that captures the structural logic of reasoning and uses two planning-oriented rewards to enhance agent learning. |
| Outcome: | The proposed method outperforms existing LLM agents on representative agent benchmarks and outperformed the current leading baselines. |
An Empirical Study of Instruction-tuning Large Language Models in Chinese (2023.findings-emnlp)
Copied to clipboard
| Challenge: | emergence of ChatGPT validates the potential of large language models (LLMs) in artificial general intelligence (AGI) however, the closed source of LLMs coupled with the requirement for massive computing resources has deterred researchers from reaching the LLM training stage. |
| Approach: | They propose to use Chinese instruction-tuning LLMs as a cookbook for customizing LLM models that can better respond to Chinese instructions. |
| Outcome: | The proposed LLM can be used to customize Chinese LLMs that can better respond to Chinese instructions. |
Can Public Large Language Models Help Private Cross-device Federated Learning? (2024.findings-naacl)
Copied to clipboard
| Challenge: | Recent studies have shown that public data can be used to improve privacy-utility trade-offs for large and small language models. |
| Approach: | They propose to use large-scale public data to help differentially private FL training . they propose a distribution matching algorithm with theoretical grounding to sample public data close to private data distribution . |
| Outcome: | The proposed method is efficient and effective for training private models by taking advantage of public data. |
Multi-Programming Language Sandbox for LLMs (2025.acl-demo)
Copied to clipboard
Shihan Dou, Jiazheng Zhang, Jianxiang Zang, Yunbo Tao, Weikang Zhou, Haoxiang Jia, Shichun Liu, Yuming Yang, Shenxi Wu, Zhiheng Xi, Muling Wu, Rui Zheng, Changze Lv, Limao Xiong, Shaoqing Zhang, Lin Zhang, Wenyu Zhan, Rongxiang Weng, Jingang Wang, Xunliang Cai, Yueming Wu, Ming Wen, Yixin Cao, Tao Gui, Xipeng Qiu, Qi Zhang, Xuanjing Huang
| Challenge: | MPLSandbox is an out-of-the-box multi-programming language sandbox designed to provide unified and comprehensive feedback from compiler and analysis tools for Large Language Models (LLMs). |
| Approach: | They propose a multi-programming language sandbox that provides unified feedback from compilers and analysis tools for Large Language Models. |
| Outcome: | The proposed multi-language sandbox can provide comprehensive feedback from compilers and analysis tools for large language models (LLMs). |
STAR: SQL Guided Pre-Training for Context-dependent Text-to-SQL Parsing (2022.findings-emnlp)
Copied to clipboard
Zefeng Cai, Xiangyu Li, Binyuan Hui, Min Yang, Bowen Li, Binhua Li, Zheng Cao, Weijie Li, Fei Huang, Luo Si, Yongbin Li
| Challenge: | Extensive experiments show that STAR outperforms previous pre-training methods and ranks first on the leaderboard . text-to-SQL parsing aims to translate natural language (NL) questions into executable SQL queries . |
| Approach: | They propose a SQL guided pre-training framework STAR for context-dependent text-to-SQL parsing . they propose two objectives that explore context-dependence of NL utterances and SQL queries . |
| Outcome: | The proposed framework outperforms existing methods on two downstream benchmarks and ranks first on the leaderboard. |
Critic-CoT: Boosting the Reasoning Abilities of Large Language Model via Chain-of-Thought Critic (2025.findings-acl)
Copied to clipboard
Xin Zheng, Jie Lou, Boxi Cao, Xueru Wen, Yuqiu Ji, Hongyu Lin, Yaojie Lu, Xianpei Han, Debing Zhang, Le Sun
| Challenge: | Existing approaches to improve the reasoning performance of large language models rely on intuitive instance-level feedback, which limits the reasoning capabilities. |
| Approach: | They propose a framework that pushes LLMs toward System-2-like critic capability by using a step-wise CoT reasoning paradigm and automatic construction of weak-supervision data without human annotation. |
| Outcome: | The proposed model significantly improves task-solving performance by filtering out invalid solutions or iterative refinement. |
Divide, Conquer, and Combine: Mixture of Semantic-Independent Experts for Zero-Shot Dialogue State Tracking (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to enhance the zeroshot generalization of DST fail to effectively decouple semantics of samples, limiting the zero-shot performance of the system. |
| Approach: | They propose a new learning schema that explicitly disentangles the semantics of seen data and leverages the performance and robustness with the mixture-of-experts mechanism. |
| Outcome: | The proposed model achieves state-of-the-art on multiWOZ2.1 with 10M trainable parameters and is robust to the mixture-of experts mechanism. |
Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for predicting hallucinations suffer from two drawbacks: Lack of scalable token-level rewards and Neglect of visual-anchored tokens. |
| Approach: | They propose a Token Preference Optimization model with self-calibrated rewards . they propose based on visual-anchored tokens and visual-aware training objective . |
| Outcome: | The proposed model improves hallucination performance by focusing on visual-anchored tokens without fine-grained annotations. |
DialogUSR: Complex Dialogue Utterance Splitting and Reformulation for Multiple Intent Detection (2022.findings-emnlp)
Copied to clipboard
Haoran Meng, Zheng Xin, Tianyu Liu, Zizhen Wang, He Feng, Binghuai Lin, Xuemin Zhao, Yunbo Cao, Zhifang Sui
| Challenge: | DialogUSR is a plug-in and domain-agnostic module that empowers multi-intent detection for chatbots . a single user query triggers inquiries on highspeed train ticket price and weather of destination. |
| Approach: | They propose a dialog utterance splitting and reformulation task that splits multi-intent user query into multiple single-intention sub-queries and recovers all coreferred and omitted information in the sub-questions. |
| Outcome: | The proposed model can be used to split multi-intent user queries into multiple sub-queries . it can be trained in two stages and perform in-depth analyses on the proposed models . |
Not All Contexts Are Equal: Teaching LLMs Credibility-aware Generation (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing RAG paradigms suffer from the impact of flawed information introduced during the retrieval phrase, thereby diminishing the reliability and correctness of the generated output. |
| Approach: | They propose a framework that empowers models to discern and process information based on its credibility. |
| Outcome: | The proposed framework outperforms existing models with retrieval augmentation and exhibits robustness despite increasing noise in the context. |
See the World, Discover Knowledge: A Chinese Factuality Evaluation for Large Vision Language Models (2025.findings-acl)
Copied to clipboard
Jihao Gu, Yingyao Wang, Pi Bu, Chen Wang, Ziming Wang, Tengtao Song, Donglai Wei, Jiale Yuan, Yingxiu Zhao, Yancheng He, Shilong Li, Jiaheng Liu, Meng Cao, Jun Song, Yingshui Tan, Xiang Li, Wenbo Su, Xiaoyong Zhu, Bo Zheng
| Challenge: | Existing models for large vision language models do not fully reflect their knowledge capacity and reliability, resulting in erroneous outputs that do not align with the image content or provide answers lacking knowledge evidence. |
| Approach: | They propose a Chinese-based benchmark for visual factuality across 8 major topics and 56 subtopics and a multi-hop question construction. |
| Outcome: | The proposed model decouples visual factuality into two parts: seeing the world and discovering knowledge. |
TRIGO: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models (2023.emnlp-main)
Copied to clipboard
Jing Xiong, Jianhao Shen, Ye Yuan, Haiming Wang, Yichun Yin, Zhengying Liu, Lin Li, Zhijiang Guo, Qingxing Cao, Yinya Huang, Chuanyang Zheng, Xiaodan Liang, Ming Zhang, Qun Liu
| Challenge: | Automated theorem proving (ATP) benchmarks focus on symbolic inference but rarely involve understanding complex number combination reasoning. |
| Approach: | They propose a benchmark that requires a model to reduce a trigonometric expression with step-by-step proof and evaluates a generative LM’s reasoning ability on formulas and ability to manipulate, group, and factor number terms. |
| Outcome: | The proposed benchmark evaluates a generative LM’s reasoning ability on formulas and ability to manipulate, group, and factor number terms. |
EA-Agent: A Structured Multi-Step Reasoning Agent for Entity Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Entity alignment (EA) aims to identify entities across different knowledge graphs (KGs) that refer to the same real-world object. |
| Approach: | They propose to use large language models to integrate semantic knowledge into EA to identify entities across different knowledge graphs that refer to the same object. |
| Outcome: | The proposed agent outperforms existing methods and achieves state-of-the-art performance on three benchmark datasets. |
X-LeBench: A Benchmark for Extremely Long Egocentric Video Understanding (2025.findings-emnlp)
Copied to clipboard
Wenqi Zhou, Kai Cao, Hao Zheng, Yunze Liu, Xinyi Zheng, Miao Liu, Per Ola Kristensson, Walterio W. Mayol-Cuevas, Fan Zhang, Weizhe Lin, Junxiao Shen
| Challenge: | Existing benchmark datasets focus on short to moderately long videos, leaving a substantial gap in evaluating extensive, ultra-long egocentric video recordings. |
| Approach: | X-LeBench is a benchmark dataset designed to evaluate long egocentric video recordings . it uses a life-logging pipeline to produce realistic, coherent daily plans . |
| Outcome: | X-LeBench is a new benchmark dataset designed to evaluate long-form egocentric video understanding . the approach produces realistic, coherent daily plans aligned with real-world video data . |
Evaluating LLMs’ Assessment of Mixed-Context Hallucination Through the Lens of Summarization (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models generate coherent text and follow instructions across diverse tasks, but a critical challenge in scaling LLM applications is hallucination, where the generated content lacks factual grounding or deviates from the intended discourse context. |
| Approach: | They use summarization as a representative task to evaluate LLMs' capability in detecting mixed-context hallucinations, specifically distinguishing between factual and non-factual hallucinos. |
| Outcome: | The proposed model distinguishes between factual and non-factual hallucinations, and their performance bottlenecks. |
Towards One-to-Many Visual Question Answering (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing Visual Question Answering systems are constrained to support domain-specific questions . a model trained on a single specific domain may not be competent for real-world application. |
| Approach: | They propose a task to enable a single model to answer as many different domains of questions as possible . they break the task down into the integration of three key abilities . |
| Outcome: | The proposed model can answer as many domains of questions as possible, the authors argue . the proposed model generalizes well to three extra zero-shot datasets, and the results are published. |
Neural Label Search for Zero-Shot Multi-Lingual Extractive Summarization (2022.acl-long)
Copied to clipboard
| Challenge: | Existing methods to translate sentences to other languages using heuristics are challenging. |
| Approach: | They propose a model that learns hierarchical weights for different sets of labels and applies them to other languages to translate them. |
| Outcome: | The proposed model can translate English datasets to other languages and obtain different sets of labels again using heuristics. |
The Linguistic Connectivities Within Large Language Models (2025.findings-acl)
Copied to clipboard
Dan Wang, Boxi Cao, Ning Bian, Xuanang Chen, Yaojie Lu, Hongyu Lin, Jia Zheng, Le Sun, Shanshan Jiang, Bin Dong, Xianpei Han
| Challenge: | Recent studies have discovered notable disparities in their performance across different languages. |
| Approach: | They conduct a systematic investigation into the behaviors of large language models across 27 different languages on 3 different scenarios and reveals a Linguistic Map correlates with the richness of available resources and linguistic family relations. |
| Outcome: | The proposed model demonstrates that there are significant disparities in performance across languages across 27 different languages on 3 different scenarios. |
DiffusEmp: A Diffusion Model-Based Framework with Multi-Grained Control for Empathetic Response Generation (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to generate empathetic responses are monotonous and generic, resulting in shallow empathy and few connections to the context. |
| Approach: | They propose to use explicit control to guide the empathy expression and a framework DiffusEmp to unify the utilization of dialogue context and attribute-oriented control signals. |
| Outcome: | The proposed framework outperforms baselines on EmpatheticDialogue in terms of controllability, informativeness, diversity, and diversity without the loss of context-relatedness. |
Difference-aware Knowledge Selection for Knowledge-grounded Conversation Generation (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing knowledge selection models are limited by the context, but the difference between selected knowledge at different turns is often overlooked. |
| Approach: | They propose a difference-aware knowledge selection method that computes the difference between the candidate knowledge sentences provided at the current turn and those chosen in the previous turns. |
| Outcome: | The proposed method outperforms the state-of-the-art methods in a knowledge-grounded dialog. |
Retrieve-and-Sample: Document-level Event Argument Extraction via Hybrid Retrieval Augmentation (2023.acl-long)
Copied to clipboard
| Challenge: | Recent studies show the effectiveness of retrieval augmentation in many generative NLP tasks. |
| Approach: | They investigate retrieval settings from the input and label distribution views . they further augment document-level EAE with pseudo demonstrations sampled from event semantic regions . |
| Outcome: | The proposed methods can augment document-level EAE with pseudo demonstrations . the methods can be used in generative NLP tasks such as dialogue response generation . |
RetroMAE: Pre-Training Retrieval-oriented Language Models Via Masked Auto-Encoder (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for dense retrieval are not effective, but there are still challenges. |
| Approach: | They propose a retrieval oriented pre-training paradigm based on Masked Auto-Encoder (MAE) where the sentence embedding is generated from the encoder’s masked input and the original sentence is recovered based upon the sentence embedded and decoded input via mangled language modeling. |
| Outcome: | The proposed model significantly improves the SOTA performance on a wide range of NLP benchmarks, like BEIR and MS MARCO. |
Sibyl: Empowering Empathetic Dialogue Generation in Large Language Models via Sensible and Visionary Commonsense Inference (2025.coling-main)
Copied to clipboard
Lanrui Wang, Jiangnan Li, Chenxu Yang, Zheng Lin, Hongyin Tang, Huan Liu, Yanan Cao, Jingang Wang, Weiping Wang
| Challenge: | Recent studies have focused on integrating commonsense knowledge into chatbots to enhance their ability to understand and generate dialogue responses. |
| Approach: | They propose a framework that integrates commonsense knowledge into chatbots to enable them to elicit more empathetic responses. |
| Outcome: | The proposed framework enables LLMs to uncover the implicit requirements of the conversation, aiming to elicit more empathetic responses. |
Deep Differential Amplifier for Extractive Summarization (2021.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to extract summary from document with a disproportionate ratio of selected and unselected sentences are far from human performance. |
| Approach: | They propose a model that rebalances sentence-level extractive summarization by amplifying the semantic difference between each sentence and all other sentences and applying the residual unit as the second item of the differential amplifier to deepen the architecture. |
| Outcome: | The proposed model performs competitively against state-of-the-art methods on two benchmark datasets. |
Geoparsing: Diagram Parsing for Plane and Solid Geometry with a Unified Formal Language (2026.findings-acl)
Copied to clipboard
Peijie Wang, Ming-Liang Zhang, Jun Cao, Chao Deng, Dekang Ran, Pi Bu, Hongda Sun, Xuan Zhang, Yingyao Wang, Jun Song, Bo Zheng, Fei Yin, Cheng-Lin Liu
| Challenge: | Recent advances in Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities across various vision reasoning tasks. |
| Approach: | They propose a unified formal language that integrates plane and solid geometry, comprehensively covering geometric structures and semantic relations. |
| Outcome: | The proposed language achieves state-of-the-art parsing performance and significantly boosts MLLMs’ capabilities for downstream geometry reasoning tasks. |
Revisiting the Evaluation of End-to-end Event Extraction (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing EE research uses the role-averaged evaluation metric, but it is misleading to downstream applications. |
| Approach: | They propose two new evaluation metrics that explicitly penalize wrongly identified event arguments. |
| Outcome: | The proposed evaluation metrics improve the initial evaluation by 10% . the proposed training scheme is better than the existing one, the authors show . |
MoleculeQA: A Dataset to Evaluate Factual Accuracy in Molecular Comprehension (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models generate erroneous information and evaluations fail to assess factual correctness of models. |
| Approach: | They propose to use MoleculeQA to evaluate molecular factual correctness in large language models by organizing molecules into a taxonomy and building QA pairs through human and LLM efforts. |
| Outcome: | The proposed model improves the factual correctness of generated information and enables the development of new models. |
MetaTS: Meta Teacher-Student Network for Multilingual Sequence Labeling with Minimal Supervision (2021.emnlp-main)
Copied to clipboard
| Challenge: | Sequence labeling aims to predict fine-grained sequences of labels for text, but lack of token-level annotated data hinders the effectiveness of supervised methods. |
| Approach: | They propose a Meta Teacher-Student (MetaTS) Network to alleviate data scarcity by leveraging large multilingual unlabeled data. |
| Outcome: | The proposed meta learning method alleviates data scarcity by leveraging large multilingual unlabeled data. |
Towards Robust Visual Question Answering: Making the Most of Biased Samples via Contrastive Learning (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies have shown that biased samples can be brittle for VQA models . however, the improvements on OOD data severely sacrifice the performance on the in-distribution (ID) data. |
| Approach: | They propose a contrastive learning approach that exploits biased samples for unbiased information that contributes to reasoning. |
| Outcome: | The proposed method achieves competitive performance on the OOD dataset while maintaining robustness on the ID dataset. |
Breaking the Reasoning Barrier A Survey on LLM Complex Reasoning through the Lens of Self-Evolution (2025.findings-acl)
Copied to clipboard
Tao He, Hao Li, Jingchang Chen, Runxuan Liu, Yixin Cao, Lizi Liao, Zihao Zheng, Zheng Chu, Jiafeng Liang, Ming Liu, Bing Qin
| Challenge: | OpenAI's O1 and subsequent projects like DeepSeek R1 have significantly advanced research on complex reasoning in LLMs. |
| Approach: | They analyze existing reasoning studies from the perspective of self-evolution and summarize O1-like works from open-source projects like DeepSeek R1 and Kimi-k1.5. |
| Outcome: | The proposed models are based on open-source models and pioneer advanced methodologies like Scaling Reinforcement Learning (RL). |
Learning to Evolve: A Self-Improving Framework for Multi-Agent Systems via Textual Parameter Graph Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for designing and optimizing multi-agent systems are static and do not learn from experience. |
| Approach: | They propose a framework that enables a multi-agent system to learn to evolve . they use "textual gradients" to pinpoint failures and suggest granular modifications . |
| Outcome: | a new framework enables a multi-agent system to learn to evolve . it learns from historical optimization experiences to improve its performance . |
Measuring Risk of Bias in Biomedical Reports: The RoBBR Benchmark (2025.emnlp-main)
Copied to clipboard
Jianyou Wang, Weili Cao, Longtian Bao, Youze Zheng, Gil Pasternak, Kaicheng Wang, Xiaoyue Wang, Ramamohan Paturi, Leon Bergen
| Challenge: | Systematic reviews should take into account the quality of available evidence, placing more weight on studies that use a valid methodology. |
| Approach: | They propose to use a risk-of-bias framework to assess the methodological strength of biomedical papers by combining expert reviewers' judgments with research paper sentences. |
| Outcome: | The proposed system measures the methodological strength of biomedical papers using the risk-of-bias framework used for systematic reviews. |
Coarse-to-Fine: Hierarchical Multi-task Learning for Natural Language Understanding (2022.coling-1)
Copied to clipboard
Zhaoye Fei, Yu Tian, Yongkang Wu, Xinyu Zhang, Yutao Zhu, Zheng Liu, Jiawen Wu, Dejiang Kong, Ruofei Lai, Zhao Cao, Zhicheng Dou, Xipeng Qiu
| Challenge: | Existing methods to learn downstream tasks by stitches skill block lack rationality and interpretation. |
| Approach: | They propose a hierarchical framework with a coarse-to-fine paradigm for generalized text representations from the large-scale corpus. |
| Outcome: | The proposed model learns basic language properties from all tasks and boosts performance on relevant tasks. |
STAIR: Learning Sparse Text and Image Representation in Grounded Tokens (2023.emnlp-main)
Copied to clipboard
Chen Chen, Bowen Zhang, Liangliang Cao, Jiguang Shen, Tom Gunter, Albin Jose, Alexander Toshev, Yantao Zheng, Jonathon Shlens, Ruoming Pang, Yinfei Yang
| Challenge: | State-of-the-art contrastive learning models like CLIP and ALIGN are less interpretable and suffer from inferior accuracy than dense representations. |
| Approach: | They extend CLIP and ALIGN models to build a sparse semantic representation that is interpretable and easy to integrate with existing retrieval systems. |
| Outcome: | The proposed model outperforms CLIP and ALIGN models on image and text retrieval tasks with a 4.9% and +4.3% improvement on COCO-5k textimage and imagetext retrieval respectively. |
CGoDial: A Large-Scale Benchmark for Chinese Goal-oriented Dialog Evaluation (2022.emnlp-main)
Copied to clipboard
| Challenge: | a new benchmark for goal-oriented dialog evaluation is needed to address the problem of knowledge sources, noisy user expressions, and the shortage of annotated data. |
| Approach: | They propose a Chinese benchmark for goal-oriented dialog evaluation that uses dialog sessions and 574,949 dialog turns to bridge the gap between academic benchmarks and spoken dialog scenarios. |
| Outcome: | The proposed benchmark contains 96,763 dialog sessions and 574,949 dialog turns totally. |
PIG: Privacy Jailbreak Attack on LLMs via Gradient-based Iterative In-Context Optimization (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods to evaluate privacy leakage in LLMs use memorized prefixes or simple instructions to extract data, which well-aligned models can easily block. |
| Approach: | They propose a framework targeting Personally Identifiable Information (PII) that uses in-context learning to build a privacy context and iteratively updates it with three gradient-based strategies to elicit target PII. |
| Outcome: | The proposed framework outperforms baseline methods and achieves state-of-the-art (SoTA) results on four white-box and two black-box LLMs. |
Reinforcing Agentic Search Via Reward Density Optimization (2026.acl-long)
Copied to clipboard
| Challenge: | Reinforcement Learning with Verifiable Rewards (RLVR) is a promising approach for enhancing agentic search, but its performance is often hindered by reward sparsity . |
| Approach: | They propose a new research problem to improve the reward obtained per unit of exploration cost by using a system that decomposes long-horizon tasks into intermediate objectives and assigns process-level rewards to provide denser learning signals. |
| Outcome: | The proposed framework outperforms strong baselines on several agentic search benchmarks and achieves comparable performance to that of advanced proprietary models. |
Enabling Agents to Communicate Entirely in Latent Space (2026.acl-long)
Copied to clipboard
Zhuoyun Du, Runze Wang, Huiyu Bai, Zouying Cao, Xiaoyong Zhu, Yu Cheng, Bo Zheng, Wei Chen, Haochao Ying
| Challenge: | Natural language is the de facto communication medium for LLM-based agents, but it presents a fundamental constraint . natural language downsampling limits the depth and nuance of information that can be transmitted . et al.: inter-agent latent space communication is a promising paradigm for solving complex tasks . |
| Approach: | They propose a paradigm that leverages the last hidden states of an LLM as a representation of its thought for direct communication. |
| Outcome: | The proposed paradigm outperforms fine-tuned chain-of-thought prompting and single-agent baselines even across heterogeneous models. |
Linguistic Rules-Based Corpus Generation for Native Chinese Grammatical Error Correction (2022.findings-emnlp)
Copied to clipboard
Shirong Ma, Yinghui Li, Rongyi Sun, Qingyu Zhou, Shulin Huang, Ding Zhang, Li Yangning, Ruiyang Liu, Zhongli Li, Yunbo Cao, Haitao Zheng, Ying Shen
| Challenge: | Chinese Grammatical Error Correction (CGEC) is a challenging NLP task and a common application in human daily life. |
| Approach: | They propose a linguistic rules-based approach to construct large-scale CGEC training corpora with automatically generated grammatical errors. |
| Outcome: | The proposed method improves performance of existing CGEC models and the benchmark is excellent resource for further development. |
Learning to Win Lottery Tickets in BERT Transfer via Task-agnostic Mask Training (2022.naacl-main)
Copied to clipboard
| Challenge: | Recent studies show pre-trained language models contain matching subnetworks that have similar transfer learning performance as the original PLM. |
| Approach: | They propose to prune matching subnetworks using magnitude-based pruning . they propose to optimize the subnetwork structure towards the pre-training objectives . |
| Outcome: | The proposed method is more efficient in searching subnetworks and advantageous when fine-tuning within a range of data scarcity. |
CLIO: Role-interactive Multi-event Head Attention Network for Document-level Event Extraction (2022.coling-1)
Copied to clipboard
| Challenge: | Existing methods for document-level event extraction struggle due to two intrinsic challenges: nested arguments and multiple events. |
| Approach: | They propose a role-interactive multi-event head attention network to solve two challenges . they map different events to multiple subspaces and then determine whether the current event exists . |
| Outcome: | The proposed model improves on two widely used DEE datasets on the Internet. |
Learning from the Dictionary: Heterogeneous Knowledge Guided Fine-tuning for Chinese Spell Checking (2022.findings-emnlp)
Copied to clipboard
Yinghui Li, Shirong Ma, Qingyu Zhou, Zhongli Li, Li Yangning, Shulin Huang, Ruiyang Liu, Chao Li, Yunbo Cao, Haitao Zheng
| Challenge: | Chinese Spell Checking (CSC) aims to detect and correct Chinese spelling errors. |
| Approach: | They propose a framework which renders Chinese Spell Checking model to learn heterogeneous knowledge from the dictionary in terms of phonetics, vision, and meaning. |
| Outcome: | The proposed framework renders the CSC model to learn heterogeneous knowledge from the dictionary in terms of phonetics, vision, and meaning. |
ACE-M3: Automatic Capability Evaluator for Multimodal Medical Models (2025.coling-main)
Copied to clipboard
| Challenge: | Existing metrics for multimodal large language models only focus on token overlap and may not align with human judgment. |
| Approach: | They propose an open-source model that assesses the question answering abilities of multimodal large language models. |
| Outcome: | Experiments show that the ACE-M3 model performs better than existing models and is more reliable than existing metrics. |
SPACE-2: Tree-Structured Semi-Supervised Contrastive Pre-training for Task-Oriented Dialog Understanding (2022.coling-1)
Copied to clipboard
| Challenge: | Existing methods for dialog understanding only consider self-augmented dialogs as positive samples and treat all other dialogs like negative ones. |
| Approach: | They propose a tree-structured pre-trained conversation model which learns dialog representations from limited labeled dialogs and large-scale unlabeled dialog corpora via semi-supervised contrastive pre-training. |
| Outcome: | The proposed model can achieve state-of-the-art results on the DialoGLUE benchmark. |
mPLUG: Effective and Efficient Vision-Language Learning by Cross-modal Skip-connections (2022.emnlp-main)
Copied to clipboard
Chenliang Li, Haiyang Xu, Junfeng Tian, Wei Wang, Ming Yan, Bin Bi, Jiabo Ye, He Chen, Guohai Xu, Zheng Cao, Ji Zhang, Songfang Huang, Fei Huang, Jingren Zhou, Luo Si
| Challenge: | Existing pre-trained vision-language models suffer from inefficiency and linguistic signal overwhelmed by long visual sequences in cross-modal alignment. |
| Approach: | They propose a vision-language foundation model with cross-modal skip-connections that can be pre-trained end-to-end on large-scale image-text pairs with both discriminative and generative objectives. |
| Outcome: | The proposed model achieves state-of-the-art results on a wide range of vision-language downstream tasks, including image captioning, image-text retrieval, visual grounding and visual question answering. |
TEBNER: Domain Specific Named Entity Recognition with Type Expanded Boundary-aware Network (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to label data and identify entities require large amounts of manually annotated texts for training supervised models. |
| Approach: | They propose a dictionary extension method which extracts new entities through the type expanded model. |
| Outcome: | The proposed method outperforms state-of-the-art supervised systems on different types of datasets and surpasses supervised models. |
Slot Dependency Modeling for Zero-Shot Cross-Domain Dialogue State Tracking (2022.coling-1)
Copied to clipboard
| Challenge: | Existing zero-shot learning methods ignore slot dependencies in a multidomain dialogue . experimental results show the effectiveness of our proposed method over existing state-of-art generation methods . |
| Approach: | They propose to use slot prompts combination, slot values demonstration and slot constraint object to model slot-slot dependency, slot-value dependency and slot-context dependency respectively. |
| Outcome: | The proposed method outperforms state-of-the-art methods under zero-shot/few-shot settings. |
Relaxing the Constraints: A Dual-Importance Projection Mechanism for Lifelong Model Editing (2026.findings-acl)
Copied to clipboard
Zhenghai Chen, Senbin Xu, Jiaxi Tan, Xinhua Wu, Yan Zhang, Xiawu Zheng, Shengchuan Zhang, Ke Li, Sicheng Zhao, Liujuan Cao, Rongrong Ji
| Challenge: | Existing knowledge editing methods rely on strict orthogonal projection to preserve previously edited knowledge, but this constraint limits gradient expressiveness, resulting in degradation of model generalization and overall performance as the number of edits increases. |
| Approach: | They propose a method that leverages Singular Value Decomposition to identify critical gradient subspaces and introduces a dual mechanism comprising "accumulated importance" and "projection importance" |
| Outcome: | Extensive experiments on five mainstream LLMs show that the proposed method achieves an average comprehensive performance improvement of 10.36% and effectively maintains the model’s general capabilities on downstream tasks. |
VLEU: a Method for Automatic Evaluation for Generalizability of Text-to-Image Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing metrics, such as CLIP, measure the semantic alignment between single prompts and their corresponding images, but they fail to evaluate a model’s generalizability across a broad spectrum of textual inputs. |
| Approach: | They propose a metric that leverages the power of Large Language Models to sample from the visual text domain and assess its generalizability. |
| Outcome: | The proposed metric evaluates the generalizability of T2I models and provides valuable insights during the finetuning process. |
CORN: Co-Reasoning Network for Commonsense Question Answering (2022.coling-1)
Copied to clipboard
| Challenge: | Existing work uses two independent modules to model QA content and external commonsense knowledge graph (KG) Existing research uses two separate modules to create QA contextual text representations and relationships between QA entities. |
| Approach: | They propose a commonsense question answering (QA) model that uses two independent modules to model QA contextual text representation and relationships between QA entities in KG. |
| Outcome: | The proposed model achieves state-of-the-art on QA benchmarks in the CommonsenseQA and OpenBookQA datasets. |
Non-Autoregressive Chinese ASR Error Correction with Phonological Training (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing methods to correct ASR errors focus on fixed-length corrections, but rarely consider variable-length ones. |
| Approach: | They propose a non-autoregressive method to correct Chinese ASR errors . they use phonological tokens to extend the source sentence for variable-length correction . |
| Outcome: | The proposed method improves word error rate and speeds up inference by 6.2 times compared with the autoregressive model. |
FolkScope: Intention Knowledge Graph Construction for E-commerce Commonsense Discovery (2023.findings-acl)
Copied to clipboard
Changlong Yu, Weiqi Wang, Xin Liu, Jiaxin Bai, Yangqiu Song, Zheng Li, Yifan Gao, Tianyu Cao, Bing Yin
| Challenge: | Existing intention-based studies on recommendation tasks are limited and use models to implicitly model the intention memberships. |
| Approach: | They propose a framework that leverages the generation power of large language models and human-in-the-loop annotation to semi-automatically construct the intention knowledge graph. |
| Outcome: | The proposed framework can model e-commerce knowledge and have many potential applications. |
Multi-level Adaptive Contrastive Learning for Knowledge Internalization in Dialogue Generation (2023.emnlp-main)
Copied to clipboard
Chenxu Yang, Zheng Lin, Lanrui Wang, Chong Tian, Liang Pang, Jiangnan Li, Qirong Ho, Yanan Cao, Weiping Wang
| Challenge: | Existing knowledge-grounded dialogue generation models struggle with dull and repetitive outputs, a problem commonly termed as text degeneration. |
| Approach: | They propose a framework that allows the model to "cheat" the objective by duplicating knowledge segments in a superficial pattern matching based on overlap. |
| Outcome: | The proposed framework can be applied to a WoW dataset and shows that it works across models and decoding strategies. |
Language Prior Is Not the Only Shortcut: A Benchmark for Shortcut Learning in VQA (2022.findings-emnlp)
Copied to clipboard
Qingyi Si, Fandong Meng, Mingyu Zheng, Zheng Lin, Yuanxin Liu, Peng Fu, Yanan Cao, Weiping Wang, Jie Zhou
| Challenge: | Visual Question Answering (VQA) models are prone to learn the shortcut solution formed by dataset biases rather than the intended solution. |
| Approach: | They propose a dataset that considers varying types of shortcuts by constructing different distribution shifts in multiple OOD test sets. |
| Outcome: | The proposed dataset considers varying types of shortcuts by constructing different distribution shifts in multiple OOD test sets. |
DialogVCS: Robust Natural Language Understanding in Dialogue System Upgrade (2024.naacl-long)
Copied to clipboard
Zefan Cai, Xin Zheng, Tianyu Liu, Haoran Meng, Jiaqi Han, Gang Yuan, Binghuai Lin, Baobao Chang, Yunbo Cao
| Challenge: | Existing models for natural language understanding are based on a well-defined intent 1 ontology. |
| Approach: | They propose to retrain the natural language understanding model as new data from real users are merged into existing data. |
| Outcome: | The proposed model shows that the semantically entangled intents can be recognized with an automatic workflow. |
Compressing and Debiasing Vision-Language Pre-Trained Models for Visual Question Answering (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on VQA models have found that they suffer from dataset biases and inefficient memory footprints. |
| Approach: | They investigate whether a VLP can be compressed and debiased simultaneously by searching sparse and robust subnetworks. |
| Outcome: | The proposed compression and debiasing pipelines outperform the debiased full VLPs on VQA tasks. |
Mobile-R1: Towards Interactive Capability for VLM-Based Mobile Agent via Systematic Training (2026.acl-long)
Copied to clipboard
Jihao Gu, Qihang Ai, Yingyao Wang, Pi Bu, Jingxuan Xing, Yue Cao, Zekun Zhu, Wei Jiang, Ziming Wang, Yingxiu Zhao, Ming-Liang Zhang, Jun Song, Yuning Jiang, Bo Zheng
| Challenge: | Existing approaches to training agents for visual-language models trap them in local optima, hindering exploration and error correction with the environment. |
| Approach: | They propose a hierarchical training recipe that bridges atomic action execution and strategic task completion. |
| Outcome: | The proposed training recipe bridges atomic action execution and strategic task completion. |
The Past Mistake is the Future Wisdom: Error-driven Contrastive Probability Optimization for Chinese Spell Checking (2022.findings-acl)
Copied to clipboard
Yinghui Li, Qingyu Zhou, Yangning Li, Zhongli Li, Ruiyang Liu, Rongyi Sun, Zizhen Wang, Chao Li, Yunbo Cao, Hai-Tao Zheng
| Challenge: | Chinese Spell Checking (CSC) aims to detect and correct spelling errors, which are caused by the phonological or visual similarity. |
| Approach: | They propose an Error-driven COntrastive Probability Optimization framework to refine the knowledge representations of pre-trained language models to avoid predicting common characters. |
| Outcome: | Extensive experiments and detailed analyses on SIGHAN datasets demonstrate that ECOPO is simple yet effective. |
PaperScope: A Multi-Modal Multi-Document Benchmark for Agentic Deep Research Across Massive Scientific Papers (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing benchmarks focus on single-document understanding, whereas real scientific workflows require integrating evidence from multiple papers. |
| Approach: | They propose a multi-modal multi-document benchmark for agentic deep research that integrates evidence from multiple documents. |
| Outcome: | Experimental results show that even advanced systems achieve limited scores on PaperScope . paper provides a rigorous benchmark alongside a pipeline for constructing large multi-modal, multi-source deep research datasets. |