Papers by Yutong Liu
More than Classification: A Unified Framework for Event Temporal Relation Extraction (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods for event temporal relation extraction ignore meaning of relations and wipe out their intrinsic dependency. |
| Approach: | They propose a unified event temporal relation extraction framework that transforms temporal relations into logical expressions of time points and completes the ETRE by predicting the relations between certain time points. |
| Outcome: | The proposed framework outperforms the state-of-the-art model on TB-Dense and MATRES by 0.3% on both datasets. |
AgentInit: Initializing LLM-based Multi-Agent Systems via Diversity and Expertise Orchestration for Effective and Efficient Collaboration (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing MAS initialization methods do not fully account for the collaborative needs of the generated agents in subsequent stages. |
| Approach: | They propose to use a Natural Language to Format mechanism to optimize the structure of agent teams and incorporate a natural language to format mechanism to ensure consistency and standardization. |
| Outcome: | The proposed method outperforms state-of-the-art initialization methods and pre-defined strategies across various frameworks and tasks while reducing token consumption. |
MMRC: A Large-Scale Benchmark for Understanding Multimodal Large Language Model in Real-World Conversation (2025.acl-long)
Copied to clipboard
Haochen Xue, Feilong Tang, Ming Hu, Yexin Liu, Qidong Huang, Yulong Li, Chengzhi Liu, Zhongxing Xu, Chong Zhang, Chun-Mei Feng, Yutong Xie, Imran Razzak, Zongyuan Ge, Jionglong Su, Junjun He, Yu Qiao
| Challenge: | Existing multimodal large language models lack the ability to memorize, recall, and reason in sustained interactions. |
| Approach: | They propose a multimodal real-world conversation benchmark for evaluating open-ended abilities of multimodal large language models. |
| Outcome: | The proposed benchmarks show that the models perform better in open-ended conversations. |
PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations (2026.acl-long)
Copied to clipboard
Yuhe Wu, Guangyu Wang, Yuran Chen, Jiatong Zhang, Yutong Zhang, Yujie Chen, Jiaming Shang, Guang Zhang, Zhuang Liu
| Challenge: | Existing benchmarks for hallucination evaluation rely on mixed queries and posterior evaluation, which quantifies hallucinosity severity but offers limited insight into where and why they occur. |
| Approach: | They propose a controlled benchmark that disentangles hallucinations into four dimensions: knowledge missing, knowledge errors, reasoning errors, and instruction-following errors. |
| Outcome: | The proposed model disentangles hallucinations into four dimensions: knowledge missing, knowledge errors, reasoning errors, and instruction-following errors. |
“I’ve Decided to Leak”: Probing Internals Behind Prompt Leakage Intents (2025.emnlp-main)
Copied to clipboard
Jianshuo Dong, Yutong Zhang, Liu Yan, Zhenyu Zhong, Tao Wei, Ke Xu, Minlie Huang, Chao Zhang, Han Qiu
| Challenge: | Large language models (LLMs) exhibit prompt leakage vulnerabilities, raising intellectual property and confidentiality concerns. |
| Approach: | They use probing techniques to capture LLMs’ intent-related internal representations and show that they internalize prompt leakage intents in their hidden states before generating tokens. |
| Outcome: | The proposed probes achieve 90%+ AUROC across all tested models, even when applied to new system prompts and attacks. |
AgentDropout: Dynamic Agent Elimination for Token-Efficient and High-Performance LLM-Based Multi-Agent Collaboration (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for MAS suffer from high token consumption and inefficiency due to frequent generation and communication among multiple agents. |
| Approach: | They propose a multi-agent system based on large language models that identifies redundant agents and communication across different communication rounds by optimizing the adjacency matrices of the communication graphs and eliminates them to enhance both token efficiency and task performance. |
| Outcome: | The proposed method reduces prompt token consumption and completion token consumption by 18.4% and improves task performance by 1.14. |
DeReA: Improving Idiom Translation with Detect-Retrieve-Arbitrate Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to idiom translation are limited by the constraints of static parametric memory and retrieval noise . idiomatic expressions are non-compositional units where figurative meanings diverge from literal interpretations . |
| Approach: | They propose a detect-retrieve-arbitrate framework that detects idiomatic spans by reasoning over semantic conflicts between literal and contextual meanings. |
| Outcome: | The proposed framework improves GPT-5-mini and Emerging Slang datasets on various model scales. |
Enhancing Zero-shot and Few-shot Stance Detection with Commonsense Knowledge Graph (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for stance detection are not applicable to zero-shot and few-shot scenarios. |
| Approach: | They propose a model that integrates commonsense knowledge into a stance detection model. |
| Outcome: | The proposed model outperforms state-of-the-art methods on zero-shot and few-shot stance detection tasks. |
From Scenes to Elements: Multi-Granularity Evidence Retrieval for Verifiable Multimodal RAG (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing multimodal Retrieval-Augmented Generation (RAG) systems retrieve evidence at coarse granularities, making failures unverifiable. |
| Approach: | They propose a multimodal benchmark that features real-world landmarks with annotations across multiple viewpoints and a framework that treats visual elements as first-class retrieval units through three stages: element-level detection and classification, multi-granularity cross-modal alignment for evidence retrieval, and attribution-constrained generation. |
| Outcome: | The proposed framework achieves up to 29.2% improvement over six strong baselines for this task. |
Spotlighter: Revisiting Prompt Tuning from a Representative Mining View (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Spotlighter is a lightweight token-selection framework that enhances accuracy and efficiency in prompt tuning. |
| Approach: | They propose a token-selection framework that enhances accuracy and efficiency in prompt tuning by preserving only the top-scoring tokens for downstream prediction. |
| Outcome: | The proposed framework outperforms CLIP by up to 11.19% in harmonic mean accuracy and achieves 0.8K additional FPS, with only 21 extra parameters. |
A Dual-Mind Framework for Strategic and Expressive Negotiation Agent (2025.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to negotiation dialogue focus on only one aspect, ignoring the synergistic effect of their combined synergies. |
| Approach: | They propose a dual-mind negotiation agent framework that integrates an intuitive and a deliberative module for slow, expression optimization. |
| Outcome: | The proposed framework achieves state-of-the-art on negotiation datasets showing that it improves negotiation ability. |
MemCoRL: Alternating Co-Optimization of Memory Retrieval and Utilization via Collaborative Reinforcement Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing research has proposed external memory modules for Large Language Models (LLMs) to overcome the limitations of finite input length and obtain contextual memory beyond the current input. |
| Approach: | They propose a two-stage alternating co-optimization reinforcement learning method that optimizes evidence retrieval and utilization using semantic feedback and rewards. |
| Outcome: | The proposed method outperforms baselines on lexical overlap and semantic similarity metrics, confirming the co-optimization in memory retrieval and memory utilization. |
Revisiting Commonsense Reasoning in Machine Translation: Training, Evaluation and Challenge (2023.acl-long)
Copied to clipboard
| Challenge: | CR is the ability to understand and navigate the world using basic knowledge and understanding shared by most people. |
| Approach: | They propose to incorporate pretrained knowledge into NMT models and use them as robust testbeds for investigating CR in NMT. |
| Outcome: | The proposed method improves the training of NMT models with high CR abilities and provides accurate evaluation metrics. |
Cost-Optimal Grouped-Query Attention for Long-Context Modeling (2025.emnlp-main)
Copied to clipboard
Yingfa Chen, Yutong Wu, Chenyang Song, Zhen Leng Thai, Xingyu Shen, Xu Han, Zhiyuan Liu, Maosong Sun
| Challenge: | Current GQA configurations overlook how context length influences inference cost . |
| Approach: | They propose a recipe for deriving cost-optimal GQA configurations that decouple the total head size from the hidden size and allow more flexible control over attention FLOPs. |
| Outcome: | The proposed configurations reduce memory usage and FLOPs by more than 50% compared to Llama-3's GQA, with *no degradation in model capabilities*. |
MASSW: A New Dataset and Benchmark Tasks for AI-Assisted Scientific Workflows (2025.findings-naacl)
Copied to clipboard
Xingjian Zhang, Yutong Xie, Jin Huang, Jinge Ma, Zhaoying Pan, Qijia Liu, Ziyang Xiong, Tolga Ergen, Dongsub Shim, Honglak Lee, Qiaozhu Mei
| Challenge: | Scientific innovation is driven by detailed workflows, which include critical steps such as contextualizing literature, generating ideas, validating ideas, and planning new research. |
| Approach: | They propose to use large language models to extract five key aspects from scientific publications to optimize scientific workflows. |
| Outcome: | The proposed dataset includes more than 152,000 peer-reviewed publications from 17 leading computer science conferences spanning the past 50 years. |
A Fine-grained Chinese Software Privacy Policy Dataset for Sequence Labeling and Regulation Compliant Identification (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing datasets that ignore law requirements are limited to English. |
| Approach: | They construct a Chinese privacy policy dataset that can be used to analyze software privacy policies. |
| Outcome: | The proposed dataset includes 483 Chinese Android privacy policies, over 11K sentences, and 52K fine-grained annotations. |
DiagnosisArena: Benchmarking Diagnostic Reasoning for Large Language Models (2026.findings-acl)
Copied to clipboard
Yakun Zhu, Zhongzhen Huang, Linjie Mu, Yutong Huang, Wei Nie, Jiaji Liu, Shaoting Zhang, Pengfei Liu, Xiaofan Zhang
| Challenge: | Existing medical benchmarks for diagnostic reasoning are limited in their ability to perform complex tasks. |
| Approach: | They propose to benchmark diagnostic capabilities of large language models to assess their accuracy and generalization bottlenecks. |
| Outcome: | The proposed model achieves 45.82%, 31.09%, and 17.79% accuracy, compared to current models, o3-mini, e1 and DeepSeek-R1 . |
SGIC: A Self-Guided Iterative Calibration Framework for RAG (2025.acl-long)
Copied to clipboard
| Challenge: | Existing studies on retrieval-augmented generation (RAG) focus on extracting relevant documents or refinement of specialized instructions. |
| Approach: | They propose a framework that provides LLMs with specific cues to improve their calibration efficacy . they propose an iterative self-calibration training set that harnesses uncertainty scores . |
| Outcome: | The proposed framework significantly improves performance on both closed-source and open-source LLMs. |
uMedSum: A Unified Framework for Clinical Abstractive Summarization (2025.acl-long)
Copied to clipboard
Aishik Nagar, Yutong Liu, Andy T. Liu, Viktor Schlegel, Vijay Prakash Dwivedi, Arun-Kumar Kaliya-Perumal, Guna Pratheep Kalanchiam, Yili Tang, Robby T. Tan
| Challenge: | Clinical abstractive summarization struggles to balance faithfulness and informativeness, sacrificing key information or introducing confabulations. |
| Approach: | They develop a modular hybrid framework that integrates confabulation removal and key information addition into abstractive summarization methods. |
| Outcome: | The proposed framework outperforms state-of-the-art abstractive summarization methods in both quantitative metrics and expert evaluations. |
TLUE: A Tibetan Language Understanding Evaluation Benchmark (2025.emnlp-main)
Copied to clipboard
Fan Gao, Cheng Huang, Yutong Liu, Nyima Tashi, Xiangxiang Wang, Thupten Tsering, Ban Ma-bao, Renzeng Duojie, Gadeng Luosang, Rinchen Dongrub, Dorje Tashi, Xiao Feng Cd, Yongbin Yu, Hao Wang
| Challenge: | Low-resource languages, like Tibetan, remain underrepresented in large language models' evaluations. |
| Approach: | They propose a Tibetan Language Understanding Evaluation Benchmark to assess LLMs' proficiency in Tibetan . they use a multi-task understanding benchmark and a safety benchmark to evaluate models . |
| Outcome: | The proposed benchmark shows that most large language models perform below the random baseline, especially in Tibetan language processing. |
Chain-of-Procedure: Hierarchical Visual-Language Reasoning for Procedural QA (2026.findings-acl)
Copied to clipboard
Guanhua Chen, Yutong Yao, Shenghe Sun, Ci-jun Gao, Shudong Liu, Lidia S. Chao, Feng Wan, Derek F. Wong
| Challenge: | Recent advances in vision-language models (VLMs) have achieved impressive results on standard image-text tasks, yet their capability in visual procedure question answering (VP-QA) remains largely unexplored. |
| Approach: | They propose a multimodal benchmark specifically designed for visual procedural reasoning that synergizes cross-modal procedure retrieval, context-aware step decomposition, and the next step prediction. |
| Outcome: | The proposed framework significantly outperforms baselines on visual procedure question answering (VP-QA) Experiments on six VLMs show that it performs better than baselines. |
TasTe: Teaching Large Language Models to Translate through Self-Reflection (2024.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to enhance LLMs' performance in machine translation are unable to fully exploit their instruction-following capabilities. |
| Approach: | They propose a framework for translating through self-reflection that involves two stages of inference . they propose to use the framework to refine LLMs' preliminary translations . |
| Outcome: | The proposed framework can produce translation outputs that match the quality of NMT systems. |
Revisiting the Reliability of Language Models in Instruction-Following (2026.acl-long)
Copied to clipboard
| Challenge: | Several benchmarks have been proposed to measure instruction-following accuracy, but these scores do not translate to reliable services in real-world use. |
| Approach: | They propose a new metric reliable@k and develop an automated pipeline to generate cousin prompts. |
| Outcome: | The proposed model can be instantiated with cousin prompts and generates high-quality cousin prompt data. |
ChatCite: LLM Agent with Human Workflow Guidance for Comparative Literature Summary (2025.coling-main)
Copied to clipboard
| Challenge: | Literature review is an indispensable step in the research process, but literature summary is challenging and time consuming. |
| Approach: | They propose an LLM agent with human workflow guidance for comparative literature summary . they use a human workflow to extract key elements from relevant literature and generate summaries . |
| Outcome: | The proposed method outperforms the CoT model in several dimensions. |
DGoT: Dynamic Graph of Thoughts for Scientific Abstract Generation (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods for generating abstracts involve collecting domain data and training corresponding models to complete the task of text summarization. |
| Approach: | They propose a method to train language models based on domain datasets and a Dynamic Graph of Thought (DGoT) which inherits the advantages of existing GoT prompt approach while reducing model reasoning cost. |
| Outcome: | The proposed method saves the cost of model training and improves reliability due to the hallucination problem of LLMs. |