Papers by Yilun Liu
TRN-R1-Zero: Text-rich Network Reasoning via LLMs with Reinforcement Learning Only (2026.acl-long)
Copied to clipboard
| Challenge: | Recent large language model-based approaches often overlook graph context or depend on distillation from larger models, limiting generalisation. |
| Approach: | They propose a framework for zero-shot reasoning on text-rich networks . they use a Neighbour-aware Group Relative Policy Optimisation objective . |
| Outcome: | The proposed framework optimises base LLMs using a Neighbour-aware group relative policy optimisation objective based on a novel margin gain metric for the informativeness of neighbouring signals . |
FOLIO: Natural Language Reasoning with First-Order Logic (2024.emnlp-main)
Copied to clipboard
Simeng Han, Hailey Schoelkopf, Yilun Zhao, Zhenting Qi, Martin Riddell, Wenfei Zhou, James Coady, David Peng, Yujie Qiao, Luke Benson, Lucy Sun, Alexander Wardle-Solano, Hannah Szabó, Ekaterina Zubova, Matthew Burtell, Jonathan Fan, Yixin Liu, Brian Wong, Malcolm Sailor, Ansong Ni, Linyong Nan, Jungo Kasai, Tao Yu, Rui Zhang, Alexander Fabbri, Wojciech Kryscinski, Semih Yavuz, Ye Liu, Xi Lin, Shafiq Joty, Yingbo Zhou, Caiming Xiong, Rex Ying, Arman Cohan, Dragomir Radev
| Challenge: | Existing benchmarks for logical reasoning in large language models lack language naturalness or limited complexity. |
| Approach: | They propose to use first-order logic annotations to evaluate logical reasoning capabilities of large language models. |
| Outcome: | The proposed dataset evaluates the FOL reasoning ability of supervised fine-tuning on medium-sized language models. |
P-FOLIO: Evaluating and Improving Logical Reasoning with Abundant Human-Written Reasoning Chains (2024.findings-emnlp)
Copied to clipboard
Simeng Han, Aaron Yu, Rui Shen, Zhenting Qi, Martin Riddell, Wenfei Zhou, Yujie Qiao, Yilun Zhao, Semih Yavuz, Ye Liu, Shafiq Joty, Yingbo Zhou, Caiming Xiong, Dragomir Radev, Rex Ying, Arman Cohan
| Challenge: | Existing methods on understanding the capabilities of LLMs in logical reasoning rely on binary entailment classification or synthetically derived rationales. |
| Approach: | They propose to annotate a human-annotated dataset consisting of diverse and complex reasoning chains for a set of realistic logical reasoning stories also written by humans. |
| Outcome: | The proposed model outperforms existing methods on understanding the capabilities of LLMs in logical reasoning by 10% or more. |
FinDVer: Explainable Claim Verification over Long and Hybrid-content Financial Documents (2024.emnlp-main)
Copied to clipboard
Yilun Zhao, Yitao Long, Tintin Jiang, Chengye Wang, Weiyuan Chen, Hongjun Liu, Xiangru Tang, Yiming Zhang, Chen Zhao, Arman Cohan
| Challenge: | FinDVer is a benchmark to evaluate the explainable claim verification capabilities of LLMs . financial documents are typically long, intricate and dense, and they include both quantita and numerical reasoning. |
| Approach: | They propose a benchmark to evaluate the explainable claim verification capabilities of LLMs . they assess 25 LLM systems under long-context and RAG settings . |
| Outcome: | The proposed benchmark can be used to evaluate the explainable claim verification capabilities of LLMs in financial documents. |
SPIN: Sparsifying and Integrating Internal Neurons in Large Language Models for Text Classification (2024.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have revolutionized text classification, but current paradigms rely on output of final layer . implicit internal structures that contribute to LLMs' impressive performance are neglected, forgoing potential performance gains. |
| Approach: | They propose a model-agnostic framework that sparsifies internal neurons of intermediate layers of LLMs for text classification. |
| Outcome: | The proposed framework significantly improves text classification accuracy, efficiency and interpretability. |
AbGen: Evaluating Large Language Models in Ablation Study Design and Evaluation for Scientific Research (2025.acl-long)
Copied to clipboard
Yilun Zhao, Weiyuan Chen, Zhijian Xu, Manasi Patwardhan, Chengye Wang, Yixin Liu, Lovekesh Vig, Arman Cohan
| Challenge: | a benchmark designed to evaluate the capabilities of LLMs in designing ablation studies for scientific research is available online. |
| Approach: | They propose to use a benchmark to evaluate LLMs' ability to design ablation studies . they investigate whether current automated evaluation methods are not reliable . |
| Outcome: | The benchmark compared leading LLMs with human experts on generating detailed ablation study designs . the results show that current evaluation methods are not reliable for the task . |
GDTB: Genre Diverse Data for English Shallow Discourse Parsing across Modalities, Text Types, and Domains (2024.emnlp-main)
Copied to clipboard
Yang Janet Liu, Tatsuya Aoyama, Wesley Scivetti, Yilun Zhu, Shabnam Behzad, Lauren Levine, Jessica Lin, Devika Tiwari, Amir Zeldes
| Challenge: | Existing shallow discourse parsing systems focus on the Wall Street Journal corpus, but the data is limited to the news domain and is 35 years old. |
| Approach: | They propose to use the Wall Street Journal corpus as a benchmark for PDTB-style shallow discourse parsing. |
| Outcome: | The proposed dataset is compatible with PDTB, but suffers from degradation out-of-domain. |
Parameter-Efficient Routed Fine-Tuning: Mixture-of-Experts Demands Mixture of Adaptation Modules (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing Parameter-Efficient Fine-Tuning (PEFT) strategies that focus on specialized experts are not effective for Mixture-of-Experts (MoE). |
| Approach: | They propose to integrate a dynamic routing mechanism among specialized experts in Mixture-of-Experts (MoE) . |
| Outcome: | Extensive experiments on commonsense and math reasoning tasks validate the performance and efficiency of the proposed routed approach. |
KnowledgeFMath: A Knowledge-Intensive Math Reasoning Dataset in Finance Domains (2024.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks for large language models (LLMs) are only 56.6% accurate, leaving room for improvement. |
| Approach: | They propose a benchmark to evaluate LLMs' capabilities in solving knowledge-intensive math reasoning problems using a finance-domain knowledge bank and expert-annotated solution references. |
| Outcome: | The proposed system achieves only 56.6% accuracy, leaving room for improvement. |
Clustering and Ranking: Diversity-preserved Instruction Selection through Expert-aligned Quality Estimation (2024.emnlp-main)
Copied to clipboard
Yuan Ge, Yilun Liu, Chi Hu, Weibin Meng, Shimin Tao, Xiaofeng Zhao, Mahong Xia, Zhang Li, Boxing Chen, Hao Yang, Bei Li, Tong Xiao, JingBo Zhu
| Challenge: | Existing methods for instruction data selection have limitations such as relying on fragile external APIs, being affected by biases in GPT models, or reducing the diversity of the selected instruction dataset. |
| Approach: | They propose an industrial-friendly, expert-aligned and diversity-preserved instruction data selection method: Clustering and Ranking (CaR). |
| Outcome: | The proposed method outperforms Alpaca's existing methods by 32.1% in GPT-4 evaluations. |
R2D2: Robust Data-to-Text with Replacement Detection (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to mitigate unfaithful text generation are inadequate . data-to-text generation requires a structured input format . |
| Approach: | They propose a training framework that addresses unfaithful Data-to-Text generation by training a system as a generator and faithfulness discriminator with additional replacement detection and unlikelihood learning tasks. |
| Outcome: | The proposed training framework improves FeTaQA, LogicNLG, and ToTTo fidelity on D2T systems. |
Towards Interpretable and Efficient Automatic Reference-Based Summarization Evaluation (2023.emnlp-main)
Copied to clipboard
Yixin Liu, Alexander Fabbri, Yilun Zhao, Pengfei Liu, Shafiq Joty, Chien-Sheng Wu, Caiming Xiong, Dragomir Radev
| Challenge: | Compared to neural systems, automatic metrics should be interpretable and provide intuitive insights into system performance and output quality. |
| Approach: | They propose to use a two-stage evaluation pipeline to extract basic information units from one text sequence and check the extracted units in another sequence. |
| Outcome: | The proposed metrics can provide high interpretability at both the fine-grained unit level and summary level, and one-stage metrics that achieve a balance between efficiency and interpretability. |
ReIFE: Re-evaluating Instruction-Following Evaluation (2025.naacl-long)
Copied to clipboard
Yixin Liu, Kejian Shi, Alexander Fabbri, Yilun Zhao, PeiFeng Wang, Chien-Sheng Wu, Shafiq Joty, Arman Cohan
| Challenge: | Existing evaluations of large language models (LLMs) for instruction following are incomplete. |
| Approach: | They propose to use 25 base LLMs and 15 recently proposed evaluation protocols to evaluate instruction following on 4 human-annotated datasets. |
| Outcome: | The proposed evaluations identify the best-performing base LLMs and evaluation protocols with a high degree of robustness. |
Part Represents Whole: Improving the Evaluation of Machine Translation System Using Entropy Enhanced Metrics (2022.findings-aacl)
Copied to clipboard
| Challenge: | Existing machine translation metrics have poor correlations with human assessments . entropy-based evaluations are often limited to a limited number of samples . |
| Approach: | They propose a fast and unsupervised approach to enhance machine translation metrics using entropy by introducing sentence-level difficulty. |
| Outcome: | The proposed method outperforms existing metrics on five sub-tracks in the WMT19 Metrics shared tasks. |
Evaluating Scene-based In-Situ Item Labeling for Immersive Conversational Recommendation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for evaluating item labels fail to leverage scenario-specific information modalities, present redundant information that is visually inferable, and lack latent awareness of users' information needs. |
| Approach: | They propose a principled categorization of information needs into explicit intent satisfaction and proactive information needs and define evaluation metrics for item label selection. |
| Outcome: | The proposed evaluation framework is based on IR-, LLM-, and VLM-based methods across fashion, movie recommendation, and retail shopping scenarios. |
MultiFinBen: Benchmarking Large Language Models for Multilingual and Multimodal Financial Application (2026.acl-long)
Copied to clipboard
Xueqing Peng, Lingfei Qian, Yan Wang, Ruoyu Xiang, Yueru He, Yang Ren, Mingyang Jiang, Vincent Jim Zhang, Yuqing Guo, Jeff Zhao, Huan He, Yi Han, Yun Feng, Yuechen Jiang, Yupeng Cao, Haohang Li, Yangyang Yu, Xiaoyu Wang, Penglei Gao, Shengyuan Lin, Keyi Wang, Shanshan Yang, Yilun Zhao, Zhiwei Liu, Peng Lu, Jerry Huang, Suyuchen Wang, Triantafillos Papadopoulos, Polydoros Giannouris, Efstathia Soufleri, Nuo Chen, Zhiyang Deng, Heming Fu, Yijia Zhao, Mingquan Lin, Meikang Qiu, Kaleb E Smith, Arman Cohan, Xiao-Yang Liu, Jimin Huang, Guojun Xiong, Alejandro Lopez-Lira, Xi Chen, Junichi Tsujii, Jian-Yun Nie, Sophia Ananiadou, Qianqian Xie
| Challenge: | Existing evaluations of LLMs in finance are text-only, monolingual, and largely saturated by current models. |
| Approach: | They propose a multilingual and multimodal benchmark for evaluating LLMs in real financial contexts. |
| Outcome: | The first expert-annotated multilingual and multimodal benchmark is released . it evaluates 21 leading LLMs and shows they perform better in multilingual settings . |
Physics: Benchmarking Foundation Models on University-Level Physics Problem Solving (2025.findings-acl)
Copied to clipboard
| Challenge: | a benchmark for university-level physics problem solving contains 1,297 expert-annotated problems . a proprietary model, o3-mini, achieves only 59.9% accuracy, highlighting fundamental weaknesses in scientific reasoning, conceptual understanding, and mathematical precision. |
| Approach: | They introduce Physics, a benchmark for university-level physics problem solving. |
| Outcome: | The proposed model achieves only 59.9% accuracy on the most advanced model, o3-mini . the proposed model is a powerful tool for evaluating models on advanced problems . |
LLM Safety From Within: Detecting Harmful Content with Internal Representations (2026.acl-long)
Copied to clipboard
| Challenge: | State-of-the-art guard models rely on terminal-layer representations and overlook safety-relevant features encoded across internal layers. |
| Approach: | They propose a lightweight guard model that harnesses safety neurons from LLM internals without modifying the underlying model. |
| Outcome: | The proposed model outperforms open-source guard models across multiple benchmarks while using 250 fewer trainable parameters. |
Benchmarking Generation and Evaluation Capabilities of Large Language Models for Instruction Controllable Summarization (2024.findings-naacl)
Copied to clipboard
Yixin Liu, Alexander Fabbri, Jiawen Chen, Yilun Zhao, Simeng Han, Shafiq Joty, Pengfei Liu, Dragomir Radev, Chien-Sheng Wu, Arman Cohan
| Challenge: | Recent studies have found that large language models (LLMs) can achieve state-of-the-art performance on generic summarization benchmarks, but their performance on more complex summarizing task settings is less studied. |
| Approach: | They benchmark large language models on instruction controllable text summarization . they use 4 evaluation protocols and 11 LLMs to evaluate their performance . |
| Outcome: | The proposed model performs well on instruction controllable text summarization tasks with 4 evaluation protocols and 11 LLMs. |
Judging with Many Minds: Do More Perspectives Mean Less Prejudice? On Bias Amplification and Resistance in Multi-Agent Based LLM-as-Judge (2025.findings-emnlp)
Copied to clipboard
Chiyu Ma, Enpei Zhang, Yilun Zhao, Wenjun Liu, Yaning Jia, Peijun Qing, Lin Shi, Arman Cohan, Yujun Yan, Soroush Vosoughi
| Challenge: | LLM-as-Judge frameworks provide scalable alternative to human evaluation . but the question of how intrinsic biases manifest in these settings remains unexplored . |
| Approach: | They conduct systematic analysis of four bias types in multi-agent LLM-as-Judge frameworks . they find debate framework amplifies biases sharply after initial debate . |
| Outcome: | The proposed frameworks amplify biases after debate and show they are stronger in meta-judge scenarios. |
Taming Text-to-Image Synthesis for Novices: User-centric Prompt Generation via Multi-turn Guidance (2025.emnlp-main)
Copied to clipboard
Yilun Liu, Minggui He, Feiyu Yao, Yuhe Ji, Shimin Tao, Jingzhou Du, Justin Li, Jian Gao, Zhang Li, Hao Yang, Boxing Chen, Osamu Yoshie
| Challenge: | Existing solutions for text-to-image synthesis are sensitive on textual prompts, posing a challenge for novice users. |
| Approach: | They propose a dialogue-based TIS prompt generation model that emphasizes user experience for novice users. |
| Outcome: | The proposed model emphasizes user experience for novice users . it improves user-centricity score while maintaining a competitive quality of synthesized images. |
M3SciQA: A Multi-Modal Multi-Document Scientific QA Benchmark for Evaluating Foundation Models (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing evaluation benchmarks for foundation models in understanding scientific literature focus on single-document tasks. |
| Approach: | They propose a multi-modal, multi-document scientific question answering benchmark . it uses expert-annotated questions that span 70 natural language processing paper clusters . |
| Outcome: | The proposed benchmarks underperform human experts in multi-modal reasoning and retrieval of scientific data. |
The GaoYao Benchmark: A Comprehensive Framework for Evaluating Multilingual and Multicultural Abilities of Large Language Models (2026.acl-long)
Copied to clipboard
Yilun Liu, Chunguang Zhao, Mengyao Piao, Lingqi Miao, Shimin Tao, Minggui HE, Chenxin Liu, Zhang Li, null Mahongxia, Jiaxin Guo, Chen Liu, Liqun Deng, Jiansheng Wei, Xiaojun Meng, Fanyi Du, Daimeng Wei, Yanghua Xiao
| Challenge: | Existing multilingual evaluation benchmarks neglect cultural nuances and lack language coverage in subjective tasks. |
| Approach: | They propose a framework that categorizes evaluation tasks into three cultural layers and nine cognitive sub-layers. |
| Outcome: | The proposed framework surpasses prior coverage by up to 111% on 20+ LLMs. |
A Corpus of Adpositional Supersenses for Mandarin Chinese (2020.lrec-1)
Copied to clipboard
| Challenge: | Adpositions are frequent markers of semantic relations, but they are highly ambiguous and vary significantly from language to language. |
| Approach: | They propose to annotate Chinese adpositions in a corpus with all aforementioned supersenses . they adapt a framework that defined a set of supersens according to ostensibly language-independent criteria . |
| Outcome: | The proposed corpus is the first to be broadly annotated with adposition semantics in Chinese . it shows that the supersense categories are well-suited to Chinese adepositions despite syntactic differences from English . |
Revisiting the Gold Standard: Grounding Summarization Evaluation with Robust Human Evaluation (2023.acl-long)
Copied to clipboard
Yixin Liu, Alex Fabbri, Pengfei Liu, Yilun Zhao, Linyong Nan, Ruilin Han, Simeng Han, Shafiq Joty, Chien-Sheng Wu, Caiming Xiong, Dragomir Radev
| Challenge: | Existing studies for summarization evaluation exhibit low inter-annotator agreement or lack scale. |
| Approach: | They propose a modified summarization salience protocol based on fine-grained semantic units and a robust summarizing evaluation benchmark. |
| Outcome: | The proposed protocol is based on fine-grained semantic units and allows for high inter-annotator agreement. |
QTSumm: Query-Focused Summarization over Tabular Data (2023.emnlp-main)
Copied to clipboard
Yilun Zhao, Zhenting Qi, Linyong Nan, Boyu Mi, Yixin Liu, Weijin Zou, Simeng Han, Ruizhe Chen, Xiangru Tang, Yumo Xu, Dragomir Radev, Arman Cohan
| Challenge: | Existing text generation systems that can provide accurate table summaries can facilitate more efficient access to relevant data insights. |
| Approach: | They propose a query-focused task where text generation models have to perform human-like reasoning and analysis over the given table to generate a tailored table summary. |
| Outcome: | The proposed method improves existing baselines on table-to-text generation and large language models by concatenating generated facts to the model input. |
ARCQuant: Boosting NVFP4 Quantization with Augmented Residual Channels for LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | NVFP4 supports fine-grained block isolation, 4-bit quantization errors and mixed-precision approaches . ARCQuant boosts NVFO4 performance via Augmented Residual Channels . |
| Approach: | They propose a framework that boosts NVFP4 performance via Augmented Residual Channels. |
| Outcome: | ARCQuant boosts NVFP4 performance via Augmented Residual Channels . the proposed framework achieves state-of-the-art accuracy comparable to full-precision baselines compared to FP16 . |
AMALGUM – A Free, Balanced, Multilayer English Web Corpus (2020.lrec-1)
Copied to clipboard
| Challenge: | a corpus of 4M tokens is available online with a large number of high-quality annotation layers. |
| Approach: | They propose to use a genre-balanced English web corpus with multiple annotation layers . they harness knowledge from multiple annotation layer to achieve a "better than NLP" benchmark . |
| Outcome: | The proposed corpus is genre-balanced and features high-quality automatic annotation layers. |