Papers by Lin Su

LLM as Effective Streaming Processor: Bridging Streaming-Batch Mismatches with Group Position Encoding (2025.findings-acl)

Challenge:	Existing approaches to task-oriented dialogue represent dialogue state as a dataflow graph . microsoft's SMCalFlow dataset features complex dialogues about events, weather, places, and people .
Approach:	They propose a dataflow graph-based dialogue agent that maps each user utterance to a program that extends this graph.
Outcome:	The proposed framework improves representability and predictability in natural dialogues . it uses dataflow graphs and metacomputation to map user intents to a program .

Copied to clipboard

Junlong Tong, Jinlan Fu, Zixuan Lin, Yingqi Fan, Anhao Zhao, Hui Su, Xiaoyu Shen

Challenge:	Existing methods for adapting LLMs to streaming rely on expensive re-encoding or limited scalability.
Approach:	They propose a group position encoding paradigm built on batch architectures to enhance consistency between streaming and batch modes.
Outcome:	The proposed method outperforms existing methods on cross-lingual and cross-modal tasks.

Knowledge Inheritance for Pre-trained Language Models (2022.naacl-main)

Copied to clipboard

Yujia Qin, Yankai Lin, Jing Yi, Jiajie Zhang, Xu Han, Zhengyan Zhang, Yusheng Su, Zhiyuan Liu, Peng Li, Maosong Sun, Jie Zhou

ToolRM: Towards Agentic Tool-Use Reward Modeling (2026.findings-acl)

Challenge:	Existing large-scale pre-trained language models are mainly trained from scratch individually, ignoring that many well-taught PLMs are available.
Approach:	They propose a pre-training framework called knowledge inheritance and propose auxiliary supervision to efficiently learn larger PLMs.
Outcome:	The proposed framework can be used to train large-scale language models with huge parameters and a large dataset can be adapted to domain adaptation and knowledge transfer.

Copied to clipboard

Renhao Li, Jianhong Tu, Yang Su, Yantao Liu, Fei Huang, Hamid Alinejad-Rokny, Derek F. Wong, Junyang Lin, Min Yang

Challenge:	lack of reliable reward models for tool-use tasks has limited progress toward agentic AI . recent advances in agentic artificial intelligence are driven by tool-using capabilities of large language models.
Approach:	They propose a pipeline that constructs pairwise preference data using rule-based scoring and multidimensional sampling to build lightweight reward models.
Outcome:	The proposed model outperforms existing models on tool calling tasks with higher accuracy.

Correctable-DST: Mitigating Historical Context Mismatch between Training and Inference for Improved Dialogue State Tracking (2022.emnlp-main)

Copied to clipboard

Hongyan Xie, Haoxiang Su, Shuangyong Song, Hao Huang, Bo Zou, Kun Deng, Jianghua Lin, Zhihui Zhang, Xiaodong He

Beta Distribution Guided Aspect-aware Graph for Aspect Category Sentiment Analysis with Affective Knowledge (2021.emnlp-main)

Challenge:	Existing dialogue state tracking approaches predict the dialogue state of a target turn sequentially based on the ground-truth previous dialogue state.
Approach:	They propose a method that predicts dialogue state sequentially based on previous dialogue state . they propose generating a previously “predicted” dialogue state using ground-truth previous dialogue states .
Outcome:	The proposed method achieves 67.51%, 68.24%, 70.30%, 71.38%, and 81.27% joint goal accuracy on MultiWOZ 2.0-2.4 datasets.

Copied to clipboard

Bin Liang, Hang Su, Rongdi Yin, Lin Gui, Min Yang, Qin Zhao, Xiaoqi Yu, Ruifeng Xu

Sentiment Analysis

Challenge:	Existing methods for aspect category sentiment analysis do not necessarily occur in a sentence.
Approach:	They propose a Beta Distribution-guided aspect-aware graph construction based on external knowledge . they use aspect-related words as the pivots to derive aspect-relevant weights .
Outcome:	The proposed approach outperforms the state-of-the-art methods on 6 benchmark datasets.

Temporal Scaling Law for Large Language Models (2025.emnlp-main)

Copied to clipboard

Yizhe Xiong, Xiansheng Chen, Xin Ye, Hui Chen, Zijia Lin, Haoran Lian, Zhenpeng Su, Wei Huang, Jianwei Niu, Jungong Han, Guiguang Ding

CartesianMoE: Boosting Knowledge Sharing among Experts via Cartesian Product Routing in Mixture-of-Experts (2025.naacl-long)

Challenge:	Existing studies have found that the test loss of LLMs scales as power-laws with model size, computational budget, and dataset size.
Approach:	They propose a concept of Temporal Scaling Law to study test loss of LLMs . they break down test loss into fine-grained token positions and develop a dynamic hyperbolic-law .
Outcome:	The proposed model predicts the test loss of LLMs as the training steps scale up.

Copied to clipboard

Zhenpeng Su, Xing W, Zijia Lin, Yizhe Xiong, Minxuan Lv, Guangyuan Ma, Hui Chen, Songlin Hu, Guiguang Ding

GEM: A General Evaluation Benchmark for Multimodal Tasks (2021.findings-acl)

Challenge:	Large language models (LLMs) have been attracting much attention due to their impressive performance in all kinds of downstream tasks.
Approach:	They propose a mix-of-experts model that allows the model size to grow without raising training costs.
Outcome:	The proposed model outperforms existing models in perplexity and robustness tests.

Copied to clipboard

Lin Su, Nan Duan, Edward Cui, Lei Ji, Chenfei Wu, Huaishao Luo, Yongfei Liu, Ming Zhong, Taroon Bharti, Arun Sacheti

Sentiment-Aware Word and Sentence Level Pre-training for Sentiment Analysis (2022.emnlp-main)

Challenge:	Existing datasets that focus on natural language tasks are not considered as a general evaluation benchmark for multimodal tasks.
Approach:	They present a general evaluation benchmark for multimodal tasks, GEM 1 . they compare it with existing multimodal vision-language datasets .
Outcome:	The proposed model is compared with existing vision-language datasets focusing on natural language tasks . it is the largest vision-linguistic dataset covering image-language tasks and video-language task at the same time .

Copied to clipboard

Shuai Fan, Chen Lin, Haonan Li, Zhenghao Lin, Jinsong Su, Hang Zhang, Yeyun Gong, JIan Guo, Nan Duan

Sentiment Analysis

Challenge:	Existing pre-trained language representation models (PLMs) capture sentiment information from word-level while under-considering sentence-level information.
Approach:	They propose a Sentiment-aware pre-trained language model with combined Word-level and Sentence-level Pre-training tasks that enhance the PLM’s knowledge about sentiment words.
Outcome:	The proposed model achieves state-of-the-art on various sentence-level and aspect-level sentiment classification benchmarks.

TACLR: A Scalable and Efficient Retrieval-based Method for Industrial Product Attribute Value Identification (2025.acl-long)

Copied to clipboard

Yindu Su, Huike Zou, Lin Sun, Ting Zhang, Haiyang Yang, Chen Li Yu, David Lo, Qingheng Zhang, Shuguang Han, Jufeng Chen

Green & Sustainable NLP

Challenge:	Existing methods for product attribute value identification face critical challenges . seller-provided attribute values are often incomplete or inaccurate .
Approach:	They propose a retrieval-based method that uses taxonomy-aware contrastive learning . they use product profiles and candidate values to encode and retrieve attributes based on similarity .
Outcome:	The proposed method is based on a taxonomy-aware, hard negative sampling and adaptive inference with dynamic thresholds.

AgentCPM-GUI: Building Mobile-Use Agents with Reinforcement Fine-Tuning (2025.emnlp-demos)

Copied to clipboard

Zhong Zhang, Yaxi Lu, Yikun Fu, Yupeng Huo, Shenzhi Yang, Yesai Wu, Han Si, Xin Cong, Haotian Chen, Yankai Lin, Xie Xie, Wei Zhou, Wang Xu, Zhou Su, Zhongwu Zhai, Xiaoming Liu, null Meiyudong, Jianming Xu, Hongyan Tian, Chongyi Wang, Chi Chen, Yuan Yao, Zhiyuan Liu, Maosong Sun

Deconvolution-Based Global Decoding for Neural Machine Translation (C18-1)

Challenge:	Large language model agents have enabled GUI-based automation, but their deployment is limited by noisy data, poor generalization, and lack of support for non-English GUIs.
Approach:	They propose an 8B-parameter GUI agent built for robust and efficient on-device GUI interaction.
Outcome:	The proposed GUI agent achieves promising performance on five public benchmarks and proposed Chinese benchmark CAGUI.

Copied to clipboard

Junyang Lin, Xu Sun, Xuancheng Ren, Shuming Ma, Jinsong Su, Qi Su

A Meaning-Based Statistical English Math Word Problem Solver (N18-1)

Challenge:	Existing models for Neural Machine Translation (NMT) use Recurrent Neural Network (RNN) to generate translation word by word following a sequential order.
Approach:	They propose a Neural Machine Translation (NMT) model that decodes the sequence with the guidance of its structural prediction of the target-side context.
Outcome:	The proposed model is more competitive compared with the state-of-the-art methods and reduces repetition with the instruction from the target-side context for decoding.

Copied to clipboard

Chao-Chun Liang, Yu-Shiang Wong, Yi-Chung Lin, Keh-Yih Su

Reasoning

Challenge:	Experimental results show that the proposed approach understands the meaning of each quantity in the text more.
Approach:	They propose a meaning-based approach for solving English math word problems . they analyze text, transform body and question parts into corresponding logic forms . Statistical models are proposed to select operator and operands .
Outcome:	The proposed approach outperforms existing systems on benchmark and noisy datasets.

RMTBench: Benchmarking LLMs Through Multi-Turn User-Centric Role-Playing (2025.findings-emnlp)

Copied to clipboard

Hao Xiang, Tianyi Tang, Yang Su, Bowen Yu, An Yang, Fei Huang, Yichang Zhang, Yaojie Lu, Hongyu Lin, Xianpei Han, Jingren Zhou, Junyang Lin, Le Sun

WR-One2Set: Towards Well-Calibrated Keyphrase Generation (2022.emnlp-main)

Challenge:	Existing benchmarks focus on character-centric approach and fail to reflect real-world applications.
Approach:	RMTBench is a user-centric bilingual role-playing benchmark featuring 80 diverse characters and over 8,000 dialogue rounds.
Outcome:	RMTBench features 80 diverse characters and over 8,000 dialogue rounds.

Copied to clipboard

Binbin Xie, Xiangpeng Wei, Baosong Yang, Huan Lin, Jun Xie, Xiaoli Wang, Min Zhang, Jinsong Su

Review-Instruct: A Review-Driven Multi-Turn Conversations Generation Method for Large Language Models (2025.findings-acl)

Challenge:	Experimental results show that keyphrase generation has serious calibration errors . ONE2SET generates short phrases summarizing an input document .
Approach:	They propose a paradigm for keyphrase generation that generates short phrases summarizing an input document.
Outcome:	The proposed model over-estimates tokens and makes it well-calibrated on common datasets.

Copied to clipboard

Jiangxu Wu, Cong Wang, Tianhuang Su, Lin Haozhi, JunYang JunYang, Zhangchao Zhangchao, Binqiang Pan, SongpanYang SongpanYang, Mingpeng Mingpeng, Kai Shi, Zixian Li

How Fast can BERT Learn Simple Natural Language Inference? (2021.eacl-main)

Challenge:	Existing methods for generating multi-turn dialogue data struggle to ensure both diversity and quality in instructions.
Approach:	They propose a framework that synthesizes multi-turn conversations through an iterative "Ask-Respond-Review" process involving three agent roles: a Candidate, multiple Reviewers, and a Chairman.
Outcome:	The proposed framework synthesizes multi-turn conversations through an iterative "Ask-Respond-Review" process involving three agent roles: a Candidate, multiple Reviewers, and a Chairman.

Copied to clipboard

Yi-Chung Lin, Keh-Yih Su

CoachMe: Decoding Sport Elements with a Reference-Based Coaching Instruction Generation Model (2025.acl-long)

Challenge:	Efficiency of learning of BERT is very slow due to hidden dataset bias . however, some studies show that it can learn with surface clues/patterns .
Approach:	They propose to use a simple entailment judgment case to test whether BERT can learn without hidden dataset bias.
Outcome:	The proposed case shows that BERT can learn without hidden bias without utilizing dataset bias.

Copied to clipboard

Wei-Hsin Yeh, Yu-An Su, Chih-Ning Chen, Yi-Hsueh Lin, Calvin Ku, Wenhsin Chiu, Min-Chun Hu, Lun-Wei Ku

Challenge:	Existing multimodal models for motion related tasks have shown significant progress.
Approach:	They propose a reference-based model that analyzes the differences between a learner’s motion and a physical reference under temporal and physical aspects.
Outcome:	The proposed model outperforms GPT-4o on figure skating and boxing by 31.6% and 58.3% respectively.

Training LLMs to be Better Text Embedders through Bidirectional Reconstruction (2025.emnlp-main)

Copied to clipboard

Chang Su, Dengliang Shi, Siyuan Huang, Jintao Du, Changhua Meng, Yu Cheng, Weiqiang Wang, Zhouhan Lin

SEE: Signal Embedding Energy for Quantifying Noise Interference in Large Audio Language Models (2026.acl-long)

Challenge:	Existing text embedding approaches often leverage the embeddment of the final token, typically a reserved special token such as ‘[EOS]‘.
Approach:	They propose to add a new training stage before contrastive learning to enrich the semantics of the final token embedding.
Outcome:	The proposed training stage improves performance on the Massive Text Embedding Benchmark (MTEB), achieving new state-of-the-art results across different LLM base models and scales.

Copied to clipboard

Yuanhe Zhang, Jiayu Tian, Yibo Zhang, Shilinlu Yan, Liang Lin, Zhenhong Zhou, Li Sun, Sen Su

Exploring the Capability Boundaries of LLMs in Mastering of Chinese Chouxiang Language (2026.findings-acl)

Challenge:	Existing studies on noise lack quantitative analysis and rely on intuition and empirical observation, thus failing to understand practical robustness.
Approach:	They propose a method for quantifying the impact of noise intensity on LALM inputs by using a structured activation subspace derived from the model's internal representations.
Outcome:	The proposed method outperforms existing denoising methods and demonstrates that noise is perceived more accurately than raw audio features.

Copied to clipboard

Dianqing Lin, Tian Lan, Jiali Zhu, Jiang Li, Wei Chen, Xu Liu, null Aruukhan, Xiangdong Su, Hongxu Hou, Guanglai Gao

LLM-OREF: An Open Relation Extraction Framework Based on Large Language Models (2025.emnlp-main)

Challenge:	Current state-of-the-art LLMs exhibit clear limitations on multiple tasks, while performing well on tasks that involve contextual semantic understanding.
Approach:	They propose a mouse-based benchmark to evaluate LLMs' performance on NLP tasks involving Chouxiang Language.
Outcome:	The proposed benchmark evaluates the performance of LLMs on six NLP tasks involving Chouxiang Language.

Copied to clipboard

Hongyao Tu, Liang Zhang, Yujie Lin, Xin Lin, Haibo Zhang, Long Zhang, Jinsong Su

Information Extraction & Text Mining

Challenge:	Existing studies focus on building models that can only handle predefined relations . however, their reliance on human annotation limits their practicality .
Approach:	They propose an open relation extraction framework that can generalize to new relations not encountered during training.
Outcome:	The proposed framework can generalize to new relations not encountered during training.

Exploring the Impact of Model Scaling on Parameter-Efficient Tuning (2023.emnlp-main)

Copied to clipboard

Yusheng Su, Chi-Min Chan, Jiali Cheng, Yujia Qin, Yankai Lin, Shengding Hu, Zonghan Yang, Ning Ding, Xingzhi Sun, Guotong Xie, Zhiyuan Liu, Maosong Sun

Global Encoding for Abstractive Summarization (P18-2)

Challenge:	Parameter-efficient tuning (PET) methods can drive large pre-trained language models by training only minimal parameters.
Approach:	They propose a parameter-efficient tuning method that is compatible with a tunable module and uses a random number generator to optimize fewer table parameters.
Outcome:	The proposed method is compatible with a tunable module and tested on 11 NLP tasks.

Copied to clipboard

Junyang Lin, Xu Sun, Shuming Ma, Qi Su

On Transferability of Prompt Tuning for Natural Language Processing (2022.naacl-main)

Challenge:	Existing models for abstractive summarization suffer from repetition and semantic irrelevance.
Approach:	They propose a global encoding framework which controls the information flow from the encoder to the decoder based on the global information of the source context.
Outcome:	The proposed model outperforms baseline models on the LCSTS and English Gigaword and can generate summary of higher quality and reduce repetition.

Copied to clipboard

Yusheng Su, Xiaozhi Wang, Yujia Qin, Chi-Min Chan, Yankai Lin, Huadong Wang, Kaiyue Wen, Zhiyuan Liu, Peng Li, Juanzi Li, Lei Hou, Maosong Sun, Jie Zhou

Responsible & Trustworthy NLP

Challenge:	Pre-trained language models (PLMs) can achieve comparable performance to full-parameter fine-tuning by tuning a few soft prompts, but require much more training time than fine-timing.
Approach:	They empirically investigate the transferability of soft prompts across different downstream tasks and PLMs to determine what decides prompt transferability.
Outcome:	The proposed method can achieve comparable performance to full-parameter fine-tuning by tuning a few soft prompts, but requires much more training time than fine-timing.

Embedding Dynamic Attributed Networks by Modeling the Evolution Processes (2020.coling-main)

Copied to clipboard

Zenan Xu, Zijing Ou, Qinliang Su, Jianxing Yu, Xiaojun Quan, ZhenKun Lin

Representation Learning

Challenge:	Existing methods to embed nodes into low-dimensional vectors focus on static networks, but in practice, many networks are evolving over time and hence are dynamic, e.g., social networks.
Approach:	They propose to extract high-order neighborhood information at each given timestamp and then use an embedding prediction framework to capture the temporal correlations.
Outcome:	Extensive experiments on four real-world datasets show that the proposed method outperforms baseline methods for dynamic link prediction and node classification tasks.

Specificity-Driven Cascading Approach for Unsupervised Sentiment Modification (D19-1)

Copied to clipboard

Pengcheng Yang, Junyang Lin, Jingjing Xu, Jun Xie, Qi Su, Xu Sun

Sentiment Analysis

Challenge:	Existing methods for unsupervised sentiment modification lack specific information in text generated without parallel data . specificity-driven cascading approach can improve specificity of generated text and content preservation .
Approach:	They propose a specificity-driven cascading approach for unsupervised sentiment modification . the method performs target sentiment addition and content reconstruction independently .
Outcome:	The proposed method outperforms competitive systems by a large margin on Yelp and Amazon datasets.

Multimodal Language Models See Better When They Look Shallower (2025.emnlp-main)

Copied to clipboard

Haoran Chen, Junyan Lin, Xinghao Chen, Yue Fan, Jianfeng Dong, Xin Jin, Hui Su, Jinlan Fu, Xiaoyu Shen

An Exploratory Study on Model Compression for Text-to-SQL (2023.findings-acl)

Challenge:	Existing studies show that multimodal large language models extract visual features from the final layers of a pretrained Vision Transformer.
Approach:	They propose a feature fusion method that strategically incorporates shallower layers . they propose MLLMs that extract visual features from the final layers of a pretrained Vision Transformer .
Outcome:	The proposed method outperforms deep layers on fine-grained visual tasks . it is the first comprehensive study of visual layer selection for MLLMs .

Copied to clipboard

Shuo Sun, Yuze Gao, Yuchen Zhang, Jian Su, Bin Chen, Yingzhan Lin, Shuqi Sun

Programming Languages in NLP

Challenge:	Text-to-SQL translates user queries into SQL statements that can retrieve relevant answers from relational databases.
Approach:	They propose to apply model compression techniques to sketch-based and sequence-to-sequence Text-toSQL models.
Outcome:	The proposed models have higher inference efficiency and respond better to model compression than sequence-to-sequence models.

Simplifying Neural Machine Translation with Addition-Subtraction Twin-Gated Recurrent Networks (D18-1)

Copied to clipboard

Biao Zhang, Deyi Xiong, Jinsong Su, Qian Lin, Huiji Zhang

RAVR: Reference-Answer-guided Variational Reasoning for Large Language Models (2026.findings-acl)

Challenge:	Existing gated recurrent networks have a vanishing gradient, allowing for more matrix transformations and less transparent functions.
Approach:	They propose an additionsubtraction twin-gated recurrent network (ATR) to simplify neural machine translation.
Outcome:	The proposed system is more transparent than LSTM/GRU due to the simplification.

Copied to clipboard

Tianqianjin Lin, Xi Zhao, Xingyao Zhang, Rujiao Long, Yi Xu, Zhuoren Jiang, Wenbo Su, Bo Zheng

RIVA: A Pre-trained Tweet Multimodal Model Based on Text-image Relation for Multimodal NER (2020.coling-main)

Challenge:	Experiments show that reinforcement learning (RL) can refine the reasoning abilities of large language models (LLMs) but requires a key prerequisite: the model must already be able to generate high-utility reasoning paths with non-negligible probability.
Approach:	They propose a framework that uses answer-conditioned reasoning as a variational surrogate for question-only reasoning.
Outcome:	Experiments on 11 benchmarks and 3 models show that RAVR reduces hesitation, strengthens conclusion consolidation, and promotes problem-specific strategies in reasoning.

Copied to clipboard

Lin Sun, Jiquan Wang, Yindu Su, Fangsheng Weng, Yuxuan Sun, Zengwei Zheng, Yuanyi Chen

Beyond Dialogue Time: Temporal Semantic Memory for Personalized LLM Agents (2026.findings-acl)

Challenge:	Named entity recognition (MNER) for tweets is a key task of many applications.
Approach:	They propose a pre-trained multimodal named entity recognition model based on Relationship Inference and Visual Attention (RIVA) for tweets.
Outcome:	The proposed model improves on the multimodal named entity recognition (MNER) task on tweets with the aid of visual clues.

Copied to clipboard

Miao Su, Yucan Guo, Zhongni Hou, Long Bai, Zixuan Li, Yufei Zhang, Guojun Yin, Wei Lin, Xiaolong Jin, Jiafeng Guo, Xueqi Cheng

Bridging the Domain Gaps in Context Representations for k-Nearest Neighbor Neural Machine Translation (2023.acl-long)

Challenge:	Existing methods focus on point-wise memory, losing durative information that captures persistent states and evolving patterns.
Approach:	They propose a memory framework that models semantic time for point-wise memory and supports the construction and utilization of durative memory.
Outcome:	Experiments on LongMemEval and LoCoMo show that the proposed method outperforms existing methods and achieves up to 12.2% improvement in accuracy.

Copied to clipboard

Zhiwei Cao, Baosong Yang, Huan Lin, Suhang Wu, Xiangpeng Wei, Dayiheng Liu, Jun Xie, Min Zhang, Jinsong Su

Semantic-Unit-Based Dilated Convolution for Multi-Label Text Classification (D18-1)

Challenge:	Existing methods to improve k-Nearest neighbor machine translation (kNN-MT) are based on the ability to non-parametrically adapt to new domains.
Approach:	They propose a method to boost the datastore retrieval of k-Nearest neighbor machine translation by reconstructing the original datastore.
Outcome:	The proposed method boosts the retrieval and translation quality of k-Nearest neighbor machine translation by reconstructing the original datastore.

Copied to clipboard

Junyang Lin, Qi Su, Pengcheng Yang, Shuming Ma, Xu Sun

Information Extraction & Text Mining

Challenge:	a novel model for multi-label text classification is proposed for the task of assigning multiple labels for a given text.
Approach:	They propose a novel model for multi-label text classification based on sequence-to-sequence learning and a hybrid attention mechanism that extracts both the word-level and the semantic unit.
Outcome:	The proposed model is competitive to the baseline models and more robust to classifying low-frequency labels.

Bridging the Gap between Training and Inference: Multi-Candidate Optimization for Diverse Neural Machine Translation (2022.findings-naacl)

Copied to clipboard

Huan Lin, Baosong Yang, Liang Yao, Dayiheng Liu, Haibo Zhang, Jun Xie, Min Zhang, Jinsong Su

Multilinguality

Challenge:	Existing diverse NMT models lack translation diversity due to a discrepancy between training and inference . despite the success of diverse NTM, there is still a lack of translation diversity .
Approach:	They propose a multi-candidate optimization framework for diverse NMT to deal with this defect.
Outcome:	The proposed framework is transparent to basic diverse NMT models, and universally makes better trade-off between diversity and quality.

Unveiling Narrative Reasoning Limits of Large Language Models with Trope in Movie Synopses (2024.findings-emnlp)

Copied to clipboard

Hung-Ting Su, Ya-Ching Hsu, Xudong Lin, Xiang-Qian Shi, Yulei Niu, Han-Yuan Hsu, Hung-yi Lee, Winston Hsu

DSMoE: Matrix-Partitioned Experts with Dynamic Routing for Computation-Efficient Dense LLMs (2025.emnlp-main)

Challenge:	Large language models (LLMs) equipped with chain-of-thoughts (CoT) prompting have shown significant multi-step reasoning capabilities in factual content like mathematics, commonsense, and logic.
Approach:	They introduce a trope-wise querying approach to assess the abstract reasoning abilities of large language models (LLMs) and uncover their low performance.
Outcome:	The proposed approach boosts the F1 score by 11.8 points and also reduces the performance of the large language models (LLMs) it also shows that it can cause hallucinations in narrative content, reducing the performance.

Copied to clipboard

Minxuan Lv, Zhenpeng Su, Leiyu Pan, Yizhe Xiong, Zijia Lin, Hui Chen, Wei Zhou, Jungong Han, Guiguang Ding, Wenwu Ou, Di Zhang, Kun Gai, Songlin Hu

BrowseComp-Plus: A Fair and Disentangled Evaluation Benchmark for Deep Search Agents (2026.acl-long)

Challenge:	Existing sparsification methods like pruning can lose model knowledge through parameter removal.
Approach:	They propose a novel approach that achieves sparsification by partitioning pre-trained FFN layers into computational blocks.
Outcome:	The proposed approach achieves superior performance across language modeling and downstream tasks under equivalent computational constraints.

Copied to clipboard

Zijian Chen, Xueguang Ma, Shengyao Zhuang, Ping Nie, Kai Zou, Sahel Sharifymoghaddam, Andrew Liu, Joshua Green, Kshama Patel, Ruoxi Meng, Mingyi Su, Yanxi Li, Haoran Hong, Xinyu Shi, Xuye Liu, Hosna Oyarhoseini, Nandan Thakur, Crystina Zhang, Luyu Gao, Wenhu Chen, Jimmy Lin

DeepPlanning: Benchmarking Long-Horizon Agentic Planning with Verifiable Constraints (2026.acl-long)

Challenge:	Existing benchmarks for deep search agents rely on blackbox web search APIs . dynamic and opaque web APIs hinder reproducibility and fair comparisons - authors .
Approach:	They propose a benchmark that employs a fixed corpus for controlled retrieval for deep search agents.
Outcome:	The new benchmark shows that agents that combine large language models with retrieval tools excel at complex, reasoning-intensive queries.

Copied to clipboard

Yinger Zhang, Shutong Jiang, Renhao Li, Jianhong Tu, Yang Su, Lianghao Deng, Xudong Guo, ChenXu Lv, Junyang Lin

Reasoning

Challenge:	Existing LLM planning benchmarks emphasize local, step-level reasoning rather than global constrained optimization.
Approach:	They propose a benchmark for practical long-horizon agent planning that uses local constrained reasoning and global constrained optimization.
Outcome:	The proposed benchmarks show that even frontier agentic LLMs struggle with these problems.

Contextual Domain Classification with Temporal Representations (2021.naacl-industry)

Copied to clipboard

Tzu-Hsiang Lin, Yipeng Shi, Chentao Ye, Yang Fan, Weitong Ruan, Emre Barut, Wael Hamza, Chengwei Su

Information Retrieval

Challenge:	Existing studies that incorporate context in SLU have focused on domains where context is limited to a few minutes.
Approach:	They propose temporal representations that combine wall-clock second difference and turn order offset information to utilize both recent and distant context in a novel large-scale setup.
Outcome:	The proposed model reduces 13.04% of classification errors compared to baseline . previous studies have focused on domains where context is limited to a few minutes .

Token-Level Policy Optimization: Linking Group-Level Rewards to Token-Level Aggregation via sequence-level likelihood (2026.acl-long)

Copied to clipboard

Xingyu Lin, Yilin Wen, Du Su, En Wang, Wenbin Liu, Zhonghou Lv, Jinchang Hou, Chenfu Bao

Reasoning

Challenge:	Group Relative Policy Optimization (GRPO) has significantly advanced the reasoning ability of large language models (LLMs).
Approach:	They propose a token-level framework that leverages sequence-level likelihood to link group-level rewards with individual tokens via token- level aggregation and introduces a KL-Divergence mask constraint that targets tokens with positive advantages and decreasing entropy to mitigate abrupt policy updates.
Outcome:	Experiments show that TEPO achieves state-of-the-art performance on mathematical reasoning benchmarks and reduces convergence time by 50% compared with GRPO/DAPO.

Pseudo2Real: Task Arithmetic for Pseudo-Label Correction in Automatic Speech Recognition (2026.findings-acl)

Copied to clipboard

Yi-Cheng Lin, Yu-Hsuan Li Liang, Hsuan Su, Tzu-Quan Lin, Shang-Tse Chen, Yun-Nung Chen, Hung-yi Lee

MiLe Loss: a New Loss for Mitigating the Bias of Learning Difficulties in Generative Language Models (2024.findings-naacl)

Challenge:	ASR models can be used to correct accent-specific errors without ground truth . pseudo-labels inherit the teacher model's systematic biases, authors say .
Approach:	They propose a parameter-space correction technique that captures pseudo-label biases . they propose achieving up to 35% relative WER reduction on a pseudo-labeled target model .
Outcome:	The proposed model achieves 35% relative WER reduction on ten African accents with the Whisper tiny model.

Copied to clipboard

Zhenpeng Su, Zijia Lin, Baixue Baixue, Hui Chen, Songlin Hu, Wei Zhou, Guiguang Ding, Xing W

Chinese SimpleQA: A Chinese Factuality Evaluation for Large Language Models (2025.acl-long)

Challenge:	Existing generative language models neglect an inherent challenge in text corpus during training, i.e., the imbalance between frequent tokens and infrequent ones.
Approach:	They propose a function to mitigate the imbalance between frequent and infrequent tokens . authors propose 'MiLe Loss' function to assess learning difficulty of tokens during training .
Outcome:	Experiments show that models with proposed model can improve on downstream benchmarks.

Copied to clipboard

Yancheng He, Shilong Li, Jiaheng Liu, Yingshui Tan, Weixun Wang, Hui Huang, Xingyuan Bu, Hangyu Guo, Chengwei Hu, Boren Zheng, Zhuoran Lin, Dekai Sun, Zhicheng Zheng, Wenbo Su, Bo Zheng

Entropy Ratio Clipping as a Soft Global Constraint for Stable Reinforcement Learning (2026.findings-acl)

Challenge:	Current frontier models sometimes generate false outputs or answers that are not substantiated by evidence.
Approach:	They propose Chinese SimpleQA, a Chinese benchmark to evaluate LLMs' factuality . they focus on Chinese language over 6 major topics with 99 diverse subtopics .
Outcome:	The Chinese SimpleQA benchmark evaluates the factuality ability of LLMs . the questions and answers are short and easy-to-evaluate .

Copied to clipboard

Zhenpeng Su, Leiyu Pan, Minxuan Lv, Tiehua Mei, Zijia Lin, Yuntao Li, Wenping Hu, Ruiming Tang, Kun Gai, Guorui Zhou

HCRE: LLM-based Hierarchical Classification for Cross-Document Relation Extraction with a Prediction-then-Verification Strategy (2026.findings-acl)

Challenge:	Large language model post-training often adopts an off-policy training paradigm . however, the off-poliicy training model introduces distribution shifts that push the policy beyond the trust region.
Approach:	They propose to use the entropy ratio as a global metric to measure the relative change in policy exploration throughout updates.
Outcome:	Experiments show that the proposed metric improves performance across multiple benchmarks.

Copied to clipboard

Guoqi Ma, Liang Zhang, Hongyao Tu, Hao Fu, Hui Li, Yujie Lin, Longyue Wang, Weihua Luo, Jinsong Su

Information Extraction & Text Mining

Challenge:	Existing approaches to cross-document relation extraction (RE) focus on identifying relations between head and tail entities from single sentence or document.
Approach:	They propose a hierarchical relation tree-based LLM-based hierarchic classification model for cross-document relation extraction (HCRE) based on predefined relations, the model can perform hierarchically classification level by level.
Outcome:	The proposed model outperforms existing baselines and validates its effectiveness.

Generative Semantic Hashing Enhanced via Boltzmann Machines (2020.acl-main)

Copied to clipboard

Lin Zheng, Qinliang Su, Dinghan Shen, Changyou Chen

Information Retrieval

Challenge:	Existing methods for generative semantic hashing assume a factorized posterior distribution, enforcing independence among the bits of hash codes.
Approach:	They propose to use a Boltzmann machine distribution as the variational posterior to introduce correlations among the bits of hash codes.
Outcome:	The proposed method can achieve significant performance gains by combining two hash codes.

Dialogue Response Selection with Hierarchical Curriculum Learning (2021.acl-long)

Copied to clipboard

Yixuan Su, Deng Cai, Qingyu Zhou, Zibo Lin, Simon Baker, Yunbo Cao, Shuming Shi, Nigel Collier, Yan Wang

MT-Bench-101: A Fine-Grained Benchmark for Evaluating Large Language Models in Multi-Turn Dialogues (2024.acl-long)

Challenge:	Empirical studies on three benchmark datasets with three state-of-the-art matching models demonstrate that the proposed learning framework significantly improves the model performance across various evaluation metrics.
Approach:	They propose a hierarchical curriculum learning framework that trains matching models in an “easy-to-difficult” scheme.
Outcome:	The proposed framework significantly improves the model performance across evaluation metrics on three benchmark datasets with three state-of-the-art matching models.

Copied to clipboard

Ge Bai, Jie Liu, Xingyuan Bu, Yancheng He, Jiaheng Liu, Zhanhui Zhou, Zhuoran Lin, Wenbo Su, Tiezheng Ge, Bo Zheng, Wanli Ouyang

Natural Language Interfaces

Challenge:	Large Language Models (LLMs) have greatly enhanced dialogue systems, but evaluation of their capabilities remains a challenge.
Approach:	They propose a model to evaluate the fine-grained abilities of Large Language Models in multi-turn dialogues.
Outcome:	The proposed model evaluates 21 popular chatbots based on MT-Bench-101 . it includes 3 overarching abilities and 13 distinct tasks within multi-turn dialogue scenarios.

Investigating Inference-time Scaling for Chain of Multi-modal Thought: A Preliminary Study (2025.findings-acl)

Copied to clipboard

Yujie Lin, Ante Wang, Moye Chen, Jingyao Liu, Hao Liu, Jinsong Su, Xinyan Xiao

Golden Touchstone: A Comprehensive Bilingual Benchmark for Evaluating Financial Large Language Models (2025.findings-emnlp)

Challenge:	Inference-time scaling of chain-of-thought (CoT) has been demonstrated as a promising approach for addressing multi-modal reasoning tasks.
Approach:	They propose to integrate visual and textual modalities within the reasoning process . they adopt a consistency-enhanced verifier to ensure effective guidance for both methods across different thought paradigms.
Outcome:	The proposed method outperforms text-only reasoning on 10 tasks spanning diverse domains and requires higher token consumption for processing richer visual inputs.

Copied to clipboard

Xiaojun Wu, Junxi Liu, Huan-Yi Su, Zhouchi Lin, Yiyan Qi, Chengjin Xu, Jiajun Su, Jiajie Zhong, Fuwei Wang, Saizhuo Wang, Fengrui Hua, Jia Li, Jian Guo

Nemotron-CC: Transforming Common Crawl into a Refined Long-Horizon Pretraining Dataset (2025.acl-long)

Challenge:	Existing financial benchmarks suffer from limited language and task coverage, low-quality datasets, and inadequate adaptability for LLM evaluation.
Approach:	They propose a bilingual benchmark for financial LLMs that assesses models’ language understanding and generation capabilities.
Outcome:	The proposed bilingual benchmark assesses models’ language understanding and generation capabilities.

Copied to clipboard

Dan Su, Kezhi Kong, Ying Lin, Joseph Jennings, Brandon Norick, Markus Kliegl, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro

Learning When to Concentrate or Divert Attention: Self-Adaptive Attention Temperature for Neural Machine Translation (D18-1)

Challenge:	Recent Common Crawl datasets remove 90% of data, limiting their suitability for long token horizon training.
Approach:	They propose to combine classifier ensembling, synthetic data rephrasing and heuristic filters to achieve better trade-offs between accuracy and data quantity.
Outcome:	The proposed model-based filtering improves MMLU by 5.6 over DCLM for 15T tokens . the full 6.3T token dataset matches DCLM on MMLO, but contains four times more unique real tokens than DCLM .

Copied to clipboard

Junyang Lin, Xu Sun, Xuancheng Ren, Muyu Li, Qi Su

Towards User-Driven Neural Machine Translation (2021.acl-long)

Challenge:	Neural Machine Translation models treat decoding at each time step equally with the same matrix . conventional methods treat decoder outputs at all time steps with the identical weight matrix causing inaccuracy .
Approach:	They propose a model with a mechanism to control the softness of attention by means of an attention temperature.
Outcome:	The proposed model outperforms baseline models on Chinese-English and English-Vietnamese translations.

Copied to clipboard

Huan Lin, Liang Yao, Baosong Yang, Dayiheng Liu, Haibo Zhang, Weihua Luo, Degen Huang, Jinsong Su

CLEVA: Chinese Language Models EVAluation Platform (2023.emnlp-demo)

Challenge:	a good translation should implicitly mirror user traits rather than translate the original content semantically.
Approach:	They propose a framework that captures user traits from historical inputs . they propose 'user-driven' NMT to model user behavior under a zero-shot learning fashion .
Outcome:	The proposed framework can capture user traits from historical inputs under zero-shot learning fashion.

Copied to clipboard

Yanyang Li, Jianqiao Zhao, Duo Zheng, Zi-Yuan Hu, Zhi Chen, Xiaohui Su, Yongfeng Huang, Shijia Huang, Dahua Lin, Michael Lyu, Liwei Wang

Who Wrote This Line? Evaluating the Detection of LLM-Generated Classical Chinese Poetry (2026.acl-long)

Challenge:	Large language models (LLMs) have revolutionized natural language processing.
Approach:	They propose a Chinese-based platform that assesses Chinese LLMs using a standardized workflow and a unique sampling strategy.
Outcome:	CLEVA evaluates Chinese LLMs on a standardized workflow and a competitive leaderboard with minimal coding.

Copied to clipboard

Jiang Li, Tian Lan, Shanshan Wang, Dongxing Zhang, Dianqing Lin, Guanglai Gao, Derek F. Wong, Xiangdong Su

Challenge:	a recent study shows that large language models can generate text, but they can also fabricate large amounts of false or misleading content.
Approach:	They propose a benchmark to detect LLM-generated classical Chinese poetry . they compare 12 different AI detectors to find out whether a poem is authored by AI .
Outcome:	The proposed benchmark compared 12 AI detectors with a dataset of 30,664 Chinese poems . the results highlight the limitations of current Chinese text detectors .

Ensuring Safe and High-Quality Outputs: A Guideline Library Approach for Language Models (2024.naacl-long)

Copied to clipboard

Yi Luo, Zhenghao Lin, YuHao Zhang, Jiashuo Sun, Chen Lin, Chengjin Xu, Xiangdong Su, Yelong Shen, Jian Guo, Yeyun Gong