Papers by Wang Zihan

Generative-to-Discriminative Test-Time Adaptation via Manifold-Aware Diffusion and Bayesian Distillation (2026.findings-acl)

Challenge:	Existing adaptive methods focus on a single axis, overlooking evidence need and reasoning depth are only partially correlated.
Approach:	They propose a dual-axis routing framework that separates retrieval necessity from reasoning necessity under a user-defined cost–quality trade-off.
Outcome:	The proposed framework reduces token usage and latency while improving answer quality over strong baselines.

Copied to clipboard

Boyun Zhang, Zequn Xie, Fangming Feng, Zihan Zhang, Yongbo He, Chuxin Wang, Sihang Cai, Tao Jin, Qifei Zhang

NavRAG: Generating User Demand Instructions for Embodied Navigation through Retrieval-Augmented LLM (2025.findings-acl)

Challenge:	Existing discriminative approaches suffer from "confident but wrong" failure mode, blindly adapting to OOD noise leading to error accumulation.
Approach:	They propose a TTA framework that harmonizes the robustness of generative diffusion models with the efficiency of discriminative regression networks via Bayesian Diffusion Distillation (BDD).
Outcome:	The proposed framework reduces MAE from 0.6872 to 0.5673 and boosts binary accuracy by 5.81 percentage points (reaching 57.33%) it also reduces the MAE of the MOSI to SIMS shift and achieves an 11.18-point gain over the baseline.

Copied to clipboard

Zihan Wang, Yaohui Zhu, Gim Hee Lee, Yachun Fan

Direct Prompt Optimization with Continuous Representations (2025.acl-long)

Challenge:	High-performance vision-and-language navigation models require large amounts of training data, the high cost of manual annotating has seriously hindered this field.
Approach:	They propose a retrieval-augmented generation framework that generates user demand instructions for vision-and-language navigation.
Outcome:	The proposed model achieves SOTA performance on the REVERIE benchmark.

Copied to clipboard

Yangkun Wang, Zihan Wang, Jingbo Shang

Context-DPO: Aligning Language Models for Context-Faithfulness (2025.findings-acl)

Challenge:	Existing methods for prompt optimization for language models lack extensibility and search space.
Approach:	They propose a method that integrates greedy strategies into optimization with continuous representations to address instability caused by rounding.
Outcome:	The proposed approach can improve prompt optimization performance on text classification and attack tasks, as well as models, including GPT-2, OPT, Vicuna, and LLaMA-2.

Copied to clipboard

Baolong Bi, Shaohan Huang, Yiwei Wang, Tianchi Yang, Zihan Zhang, Haizhen Huang, Lingrui Mei, Junfeng Fang, Zehao Li, Furu Wei, Weiwei Deng, Feng Sun, Qi Zhang, Shenghua Liu

ReportLogic: Evaluating Logical Quality in Deep Research Reports (2026.acl-long)

Challenge:	Context-DPO is the first alignment method specifically designed to enhance contextfaithfulness for large language models.
Approach:	They propose a benchmark that simulates Retrieval-Augmented Generation scenarios with knowledge conflicts to evaluate context-faithfulness.
Outcome:	The proposed method improves LLMs' context-faithfulness by 35% to 280% over open-source models.

Copied to clipboard

Jujia Zhao, Zhaoxin Huan, Zihan Wang, Xiaolu Zhang, Jun Zhou, Suzan Verberne, Zhaochun Ren

Robustness in NLP

Challenge:	Existing evaluation frameworks that evaluate large language models for Deep Research largely ignore this requirement.
Approach:	They propose a benchmark that quantifies report-level logical quality through a reader-centric lens of auditability.
Outcome:	The proposed model quantifies logical quality through a reader-centric lens of auditability.

On LLM-Based Scientific Inductive Reasoning Beyond Equations (2025.emnlp-main)

Copied to clipboard

Brian S. Lin, Jiaxin Yuan, Zihan Zhou, Shouli Wang, Shuo Wang, Cunliang Kong, Qi Shi, Yuxuan Li, Liner Yang, Zhiyuan Liu, Maosong Sun

Neo-Classic: A Benchmark for Evaluating Linguistic-Aesthetic Reasoning in Classical Chinese Poetry (2026.acl-long)

Challenge:	Existing research on inductive reasoning models emphasizes rule design without grounding them in specific scenarios.
Approach:	They propose to use LLMs to learn underlying patterns from limited examples in entirely new environments.
Outcome:	The proposed benchmark evaluates the inductive reasoning abilities of large language models in scientific settings.

Copied to clipboard

Han Zhang, Zihan Gu, Zhiyuan Wang, Tianyi Ma, Jiacheng Lu, Xinyan Zhang, Yuhao Wei, Cheng Hua

Self-Renewal Prompt Optimizing with Implicit Reasoning (2024.findings-emnlp)

Challenge:	Large Language Models (LLMs) achieve high accuracy on established Classical Chinese Poetry benchmarks, but it remains challenging to distinguish transferable Linguistic-Aesthetic Reasoning from reliance on familiar pre-training patterns.
Approach:	They propose a benchmark that combines a constructionist Out-of-Sample dataset with reverse understanding probes to evaluate large-scale large-format models.
Outcome:	The proposed model performs well on classical Chinese poetry benchmarks, but a performance gap persists . the model can complete famous couplets and can be used to understand a variety of texts.

Copied to clipboard

Zihan Liang, Ben Chen, Zhuoran Ran, Zihan Wang, Huangyu Dai, Yufei Ma, Dehong Gao, Xiaoyan Cai, Libin Yang

CreativeBench: Benchmarking and Enhancing Machine Creativity via Self-Evolving Challenges (2026.findings-acl)

Challenge:	Recent advances in NLP have been driven by the development of Large Language Models (LLMs).
Approach:	They propose a self-renewal approach to optimize LLM outputs to better align with human preferences without supervised fine-tuning.
Outcome:	The proposed approach improves outputs to better align with human preferences across LLMs and tasks without supervised fine-tuning.

Copied to clipboard

Zihan Wang, Lam Nguyen, Zhengyang Zhao, Mengyue Yang, Chengwei Qin, Yujiu Yang, Linyi Yang

Challenge:	Increasing saturation of web data limits further scaling of model intelligence.
Approach:	They propose a benchmark to evaluate machine creativity in code generation that combines combinatorial and exploratory creativity through reverse engineering and self-play.
Outcome:	The proposed benchmark targets combinatorial and exploratory creativity through reverse engineering and self-play.

Latent Suicide Risk Detection on Microblog via Suicide-Oriented Word Embeddings and Layered Attention (D19-1)

Copied to clipboard

Lei Cao, Huijun Zhang, Ling Feng, Zihan Wei, Xin Wang, Ningyun Li, Xiaohao He

LLM×MapReduce-V3: Enabling Interactive In-Depth Survey Generation through a MCP-Driven Hierarchically Modular Agent System (2025.emnlp-demos)

Challenge:	Existing approaches to detect suicidal ideation on social media are limited to a small group of people.
Approach:	They propose to use tree holes to embed words into microblogs to strengthen the sensibility of suicide-related lexicons and to use a two-layered attention mechanism to grasp intermittently changing points from individual's open blog streams.
Outcome:	The proposed approach can achieve over 91% accuracy with the use of suicide-oriented word embeddings and attention on a large-scale well-labelled suicide data set.

Copied to clipboard

Yu Chao, Siyu Lin, Xiaorong Wang, Zhu Zhang, Zihan Zhou, Haoyu Wang, Shuo Wang, Jie Zhou, Zhiyuan Liu, Maosong Sun

Challenge:	Generating high-quality long-form survey articles poses significant challenges to AI Agent systems.
Approach:	They propose a hierarchically modular agent system for long-form survey generation . they use atomic models to implement skeleton initialization, digest construction, and skelet refinement . human evaluations demonstrate system surpasses representative baselines .
Outcome:	The proposed system surpasses representative baselines in both content depth and length, highlighting the strength of MCP-based modular planning.

SongComposer: A Large Language Model for Lyric and Melody Generation in Song Composition (2025.acl-long)

Copied to clipboard

Shuangrui Ding, Zihan Liu, Xiaoyi Dong, Pan Zhang, Rui Qian, Junhao Huang, Conghui He, Dahua Lin, Jiaqi Wang

ClusterLLM: Large Language Models as a Guide for Text Clustering (2023.emnlp-main)

Challenge:	Creating lyrics and melodies in symbolic format requires expert knowledge of melody and an advanced understanding of lyrics.
Approach:	They introduce SongComposer, a music-specialized large language model that can create symbolic lyrics and melodies following instructions.
Outcome:	The proposed model outperforms existing models in symbolic song composition tasks.

Copied to clipboard

Yuwei Zhang, Zihan Wang, Jingbo Shang

FastGAS: Fast Graph-based Annotation Selection for In-Context Learning (2024.findings-acl)

Challenge:	Extensive experiments on 14 datasets show that ClusterLLM consistently improves clustering quality, at an average cost of $0.6 per dataset.
Approach:	They propose a text clustering framework that leverages feedback from an instruction-tuned large language model, such as ChatGPT.
Outcome:	Extensive experiments on 14 datasets show that ClusterLLM consistently improves clustering quality, at an average cost of $0.6 per dataset.

Copied to clipboard

Zihan Chen, Song Wang, Cong Shen, Jundong Li

HyperMoE: Towards Better Mixture of Experts via Transferring Among Experts (2024.acl-long)

Challenge:	Existing methods to select unlabeled examples for annotation require a long time due to their complexity, hindering their practical viability.
Approach:	They propose a graph-based selection method to efficiently identify high-quality instances while minimizing computational overhead.
Outcome:	The proposed method significantly reduces selection time and improves performance on different tasks.

Copied to clipboard

Hao Zhao, Zihan Qiu, Huijia Wu, Zili Wang, Zhaofeng He, Jie Fu

Is He Extroverted? Identifying Missing Relevant Personas for Faithful User Simulation (2026.eacl-srw)

Challenge:	Existing methods for enhancing performance through increased use of expert knowledge often result in diminishing sparsity during expert selection.
Approach:	They propose a framework that integrates the computational processes of MoE with the concept of knowledge transferring in multi-task learning.
Outcome:	The proposed framework outperforms existing methods under identical conditions concerning the number of experts.

Copied to clipboard

Weiwen Su, Yuhan Zhou, Zihan Wang, Naoki Yoshinaga, Masashi Toyoda

Natural Language Interfaces

Challenge:	Existing user simulation approaches focus on generating user-like responses in dialogue without verifying whether critical personas are supplied.
Approach:	They propose a task of identifying persona dimensions that are relevant but missing in simulating a user's reply for a given dialogue context.
Outcome:	The proposed model identifies persona dimensions that are relevant but missing in simulating a user’s response for a given dialogue context.

DeepMed: Building a Medical DeepResearch Agent via Multi-hop Med-Search Data and Turn-Controlled Agentic Training & Inference (2026.findings-acl)

Copied to clipboard

Zihan Wang, Hao Wang, Shi Feng, Xiaocui Yang, Daling Wang, Yiqun Zhang, Jinghao Lin, Xiaozhong Ji, Haihua Yang

HPT: Hierarchy-aware Prompt Tuning for Hierarchical Text Classification (2022.emnlp-main)

Challenge:	Medical reasoning models are constrained by parametric knowledge and can induce hallucinations and spurious attributions.
Approach:	They propose a model that uses a multi-hop med-search QA synthesis method to apply the DR paradigm in medical contexts.
Outcome:	The proposed model outperforms larger medical reasoning models on medical benchmarks.

Copied to clipboard

Zihan Wang, Peiyi Wang, Tianyu Liu, Binghuai Lin, Yunbo Cao, Zhifang Sui, Houfeng Wang

How Do Large Language Models Capture the Ever-changing World Knowledge? A Review of Recent Advances (2023.emnlp-main)

Challenge:	Hierarchical text classification (HTC) is a multi-label classification problem with a complex label hierarchy.
Approach:	They propose a Hierarchy-aware Prompt Tuning method to handle HTC from a multi-label perspective using a dynamic virtual template and label words that take the form of soft prompts to fuse the label hierarchy knowledge.
Outcome:	The proposed method achieves state-of-the-art performance on 3 popular HTC datasets and is adept at handling imbalance and low resource situations.

Copied to clipboard

Zihan Zhang, Meng Fang, Ling Chen, Mohammad-Reza Namazi-Rad, Jun Wang

Learn from Failure: Fine-tuning LLMs with Trial-and-Error Data for Intuitionistic Propositional Logic Proving (2024.acl-long)

Challenge:	Large language models (LLMs) are impressive in solving tasks, but they can quickly be outdated after deployment.
Approach:	They provide a review of recent advances in aligning deployed large language models with the ever-changing world knowledge.
Outcome:	The proposed models can be used to perform various tasks directly through in-context learning or for further fine-tuning for domain-specific uses.

Copied to clipboard

Chenyang An, Zhibo Chen, Qihao Ye, Emily First, Letian Peng, Jiayun Zhang, Zihan Wang, Sorin Lerner, Jingbo Shang

Debiasing Made State-of-the-art: Revisiting the Simple Seed-based Weak Supervision for Text Classification (2023.emnlp-main)

Challenge:	Recent advances in Automated Theorem Proving have shown the effectiveness of leveraging a (large) language model that generates tactics (i.e. proof steps) to search through proof states.
Approach:	They propose to use a large language model that generates tactics to search through proof states.
Outcome:	The proposed model solves more unseen theorems with lower trial searches than the current model, which only learns from failed attempts.

Copied to clipboard

Chengyu Dong, Zihan Wang, Jingbo Shang

Formulating Few-shot Fine-tuning Towards Language Model Pre-training: A Pilot Study on Named Entity Recognition (2022.findings-emnlp)

Challenge:	Recent advances in weakly supervised text classification focus on designing sophisticated methods to turn high-level human heuristics into quality pseudo-labels.
Approach:	They propose to use a seed matching-based method to generate quality pseudo-labels by deleting the seed words present in the matched input text.
Outcome:	The proposed method can be improved significantly by deleting the seed words in the matched input text with a high deletion ratio.

Copied to clipboard

Zihan Wang, Kewen Zhao, Zilong Wang, Jingbo Shang

Alignment for Efficient Tool Calling of Large Language Models (2025.emnlp-main)

Challenge:	Existing methods for fine-tuning pre-trained language models are limited . we propose a few-shot fine-uning framework for NER .
Approach:	They propose a few-shot fine-tuning framework for named entity recognition (NER) they propose three new types of tokens, "is-entity", "which-type" and "bracket"
Outcome:	The proposed framework improves on pre-trained language models on several benchmark datasets.

Copied to clipboard

Hongshen Xu, Zihan Wang, Zichen Zhu, Lei Pan, Xingyu Chen, Shuai Fan, Lu Chen, Kai Yu

Text Grafting: Near-Distribution Weak Supervision for Minority Classes in Text Classification (2024.emnlp-main)

Challenge:	Recent advances in tool learning have enabled large language models to integrate external tools, enhancing their task performance by expanding their knowledge boundaries.
Approach:	They propose a framework that combines probabilistic knowledge boundary estimation with dynamic decision-making to allow LLMs to better assess when to invoke tools based on their confidence.
Outcome:	The proposed framework shows significant improvements in tool efficiency by reducing unnecessary tool usage.

Copied to clipboard

Letian Peng, Yi Gu, Chengyu Dong, Zihan Wang, Jingbo Shang

EMCompress: Video-LLMs with Endomorphic Multimodal Compression (2026.findings-acl)

Challenge:	Recent work generates pseudo labels by mining texts similar to the class names from the raw corpus, but there is a high risk that LLMs cannot generate in-distribution data, leading to ungeneralizable classifiers.
Approach:	They propose to use LLMs to generate pseudo labels by mining masked templates from corpus . they then use state-of-the-art LLM to synthesize near-distribution texts falling into minority classes .
Outcome:	The proposed framework improves on the previous methods for extremely weak-supervised text classification.

Copied to clipboard

Zheyu Fan, Jiateng Liu, Yuji Zhang, Zihan Wang, Yi R. Fung, Manling Li, Heng Ji

Tears or Cheers? Benchmarking LLMs via Culturally Elicited Distinct Affective Responses (2026.acl-long)

Challenge:	Static, sparse frame sampling either dilutes evidence across task-irrelevant segments at significant cost or misses fine-grained temporal semantics altogether.
Approach:	They propose a novel task that compresses multimodal input while preserving answer invariance across reasonable downstream models.
Outcome:	The proposed task surpasses prior methods by 0.40 F-1 with competitive query rewriting.

Copied to clipboard

Chongyuan Dai, Yaling Shen, Zihan Gao, Jia Li, Yishun Jiang, Yaxiong Wang, Liu Liu, Zongyuan Ge, Jinpeng Hu

Sentiment Analysis

Challenge:	Culture is a fundamental determinant of human affective processing and affective perceptions are often limited by declarative knowledge or established societal customs.
Approach:	They propose a multimodal benchmark that leverages LLM-generated provisional labels to isolate cross-cultural emotional distinctions.
Outcome:	The proposed benchmark captures cross-cultural emotional distinctions and derives reliable ground-truth annotations through human evaluation.

SEEKR: Selective Attention-Guided Knowledge Retention for Continual Learning of Large Language Models (2024.emnlp-main)

Copied to clipboard

Jinghan He, Haiyun Guo, Kuan Zhu, Zihan Zhao, Ming Tang, Jinqiao Wang

Democratizing Reasoning Ability: Tailored Learning from Large Language Model (2023.emnlp-main)

Challenge:	Existing methods fail to fully exploit the knowledge embedded in models from previous tasks . Existing techniques fail to exploit the information embedded in previous tasks, resulting in a large number of replay samples to achieve good results.
Approach:	They propose a method that uses attention weights to extract knowledge from previous tasks . they use a data replay strategy to extract the knowledge from the previous tasks.
Outcome:	The proposed method achieves comparable or even better performance with only 1/10 of replayed data used by other methods.

Copied to clipboard

Zhaoyang Wang, Shaohan Huang, Yuxuan Liu, Jiahai Wang, Minghui Song, Zihan Zhang, Haizhen Huang, Furu Wei, Weiwei Deng, Feng Sun, Qi Zhang

MUSE: A Multimodal Conversational Recommendation Dataset with Scenario-Grounded User Profiles (2025.findings-acl)

Challenge:	Large language models (LLMs) exhibit impressive emergent abilities in natural language processing, but their democratization is hindered due to huge computation requirements and closed-source nature.
Approach:	They propose a tailored learning approach to distill the exclusive reasoning ability to smaller LMs to facilitate democratization.
Outcome:	The proposed approach enables the democratization of the exclusive reasoning ability by leveraging the black-box model as a reasoning teacher.

Copied to clipboard

Zihan Wang, Xiaocui Yang, YongKang Liu, Shi Feng, Daling Wang, Yifei Zhang

Demons in the Detail: On Implementing Load Balancing Loss for Training Specialized Mixture-of-Expert Models (2025.acl-long)

Challenge:	Existing research focuses solely on text, leaving a gap with practical applications.
Approach:	They propose to synthesize a multimodal conversational recommendation dataset using multimodal large language models to automatically synthesized data from 7,000 conversations in the Clothing domain.
Outcome:	The proposed dataset contains 83,148 utterances from 7,000 conversations centered around the Clothing domain.

Copied to clipboard

Zihan Qiu, Zeyu Huang, Bo Zheng, Kaiyue Wen, Zekun Wang, Rui Men, Ivan Titov, Dayiheng Liu, Jingren Zhou, Junyang Lin

Programming Languages in NLP

Challenge:	Existing Mixture-of-Experts training frameworks use a micro-batch to calculate LBL . micro-batches are restricted to a single sequence, preventing expert specialization .
Approach:	They propose to use a global-batch to loosen the load balance constraint for MoEs models . they propose to synchronize fi across micro-batches and then use it to calculate the LBL .
Outcome:	The proposed global-batch LBL improves the domain specialization of experts . the micro-battery LBL is almost at the sequence level, and the router is pushed to distribute the token evenly .

Answer-driven Deep Question Generation based on Reinforcement Learning (2020.coling-main)

Copied to clipboard

Liuyin Wang, Zihan Xu, Zibo Lin, Haitao Zheng, Ying Shen

Utilizing Local Hierarchy with Adversarial Training for Hierarchical Text Classification (2024.lrec-main)

Challenge:	Existing methods for deep question generation focus on enhancing document representations, but little attention is paid to the answer information.
Approach:	They propose a deep question generation model that makes better use of the target answer as a guidance to facilitate question generation.
Outcome:	The proposed model outperforms state-of-the-art models in automatic and human evaluations on the hotpotQA dataset.

Copied to clipboard

Zihan Wang, Peiyi Wang, Houfeng Wang

MoDULA: Mixture of Domain-Specific and Universal LoRA for Multi-Task Learning (2024.emnlp-main)

Challenge:	Hierarchical text classification (HTC) is a challenging subtask due to its complex taxonomic structure.
Approach:	They propose a local hierarchy framework that can fit in nearly all HTC models and optimize them with the local hierarchy as auxiliary information.
Outcome:	The proposed framework is effective in all scenarios and is adept at dealing with complex taxonomic hierarchies.

Copied to clipboard

Yufei Ma, Zihan Liang, Huangyu Dai, Ben Chen, Dehong Gao, Zhuoran Ran, Wang Zihan, Linbo Jin, Wen Jiang, Guannan Zhang, Xiaoyan Cai, Libin Yang

ChatGLM-Math: Improving Math Problem-Solving in Large Language Models with a Self-Critique Pipeline (2024.findings-emnlp)

Challenge:	Recent advances in open-source Large Language Models (LLMs) have achieved notable successes in natural language processing.
Approach:	They propose a Parameter Efficient Fine-Tuning paradigm for improved fine-tuning and parameter efficiency in multi-task learning.
Outcome:	The proposed model outperforms existing methods on multi-task learning while reducing training costs by over 80% without losing general capability.

Copied to clipboard

Yifan Xu, Xiao Liu, Xinghan Liu, Zhenyu Hou, Yueyan Li, Xiaohan Zhang, Zihan Wang, Aohan Zeng, Zhengxiao Du, Zhao Wenyi, Jie Tang, Yuxiao Dong

Numerical Reasoning

Challenge:	Large language models (LLMs) have shown excellent mastering of human language but struggle in real-world applications that require mathematical problem-solving.
Approach:	They propose a pipeline to train a general Math-Critique model from the LLM itself to provide feedback signals and employ rejective fine-tuning and direct preference optimization over the Llm's own generations for data collection.
Outcome:	The proposed pipeline outperforms existing LLMs that could be two times larger.

MatPlotAgent: Method and Evaluation for LLM-Based Agentic Scientific Data Visualization (2024.findings-acl)

Copied to clipboard

Zhiyu Yang, Zihan Zhou, Shuo Wang, Xin Cong, Xu Han, Yukun Yan, Zhenghao Liu, Zhixing Tan, Pengyuan Liu, Dong Yu, Zhiyuan Liu, Xiaodong Shi, Maosong Sun

Diagnosing Hidden Instabilities in Model Editing via Uncertainty Quantification (2026.acl-long)

Challenge:	Scientific data visualization is an essential process in research, but its use of large language models remains unexplored.
Approach:	They propose a model-agnostic LLM agent framework to automate scientific data visualization tasks.
Outcome:	The proposed framework improves performance of commercial and open-source models.

Copied to clipboard

Zihan Gu, TianYi Zhang, Xinyan Zhang, Zhiyuan Wang, Han Zhang, Yuhao Wei, Jiacheng Lu, Tianyi Ma, Xingsheng Zhang, Hua Zhang, Yue Hu

Challenge:	Existing methods to update large language models (LLMs) without expensive retraining are fragile under single-edit evaluation protocols.
Approach:	They propose a framework that characterizes activation-based editing as a constrained intervention on intermediate representations.
Outcome:	The proposed method reveals local knowledge conflicts invisible to existing benchmarks.

DualGuard: A Parameter Space Transformation Approach for Bidirectional Defense in Split-Based LLM Fine-Tuning (2025.acl-long)

Copied to clipboard

Zihan Liu, Yizhen Wang, Rui Wang, Sai Wu

E2LLM: Encoder Elongated Large Language Models for Long-Context Understanding and Reasoning (2025.emnlp-main)

Challenge:	Existing defense methods for large language model fine-tuning (LLM-FT) sacrifice task-specific performance under privacy constraints.
Approach:	They propose a bidirectional defense mechanism that uses a local warm-up parameter transformation to alter client-side model parameters before training.
Outcome:	The proposed defense mechanism outperforms current defense methods while maintaining task performance.

Copied to clipboard

Zihan Liao, Jun Wang, Hang Yu, Lingxiao Wei, Jianguo Li, Jun Wang, Wei Zhang

Powering Comparative Classification with Sentiment Analysis via Domain Adaptive Knowledge Transfer (2021.emnlp-main)

Challenge:	Considerable efforts have been and are still being put into increasing the context length of Large Language Models (LLMs)
Approach:	They propose an approach that divides long contexts into chunks, compresses each into soft prompts using a pretrained text encoder, and aligns these representations with a decoder-only LLM via an adapter.
Outcome:	The proposed approach outperforms 8 state-of-the-art methods in effectiveness and efficiency for document summarization and question answering, and achieves the best performance on LongBench v2 among models of comparable size.

Copied to clipboard

Zeyu Li, Yilong Qin, Zihan Liu, Wei Wang

A Survey of Large Language Models for Text-Guided Molecular Discovery: From Molecule Generation to Optimization (2026.acl-long)

Challenge:	Comparative Preference Classification (CPC) is a natural language processing task that predicts whether a preference comparison exists between two entities in a given sentence .
Approach:	They propose a sentiment analyzer that learns sentiments to individual entities via domain adaptive knowledge transfer.
Outcome:	Experiments on the CompSent-19 dataset present a significant improvement on the F1 scores over the best existing CPC approaches.

Copied to clipboard

Ziqing Wang, Kexin Zhang, Zihan Zhao, Yibo Wen, Abhishek Pandey, Han Liu, Kaize Ding

Learning from Diverse Reasoning Paths with Routing and Collaboration (2025.emnlp-main)

Challenge:	Large language models (LLMs) are introducing a paradigm shift in molecular discovery by enabling text-guided interaction with chemical spaces through natural language and symbolic notations.
Approach:	They analyze the current LLM learning paradigms to tackle four critical evaluation dimensions that have emerged as critical dimensions in recent studies.
Outcome:	The proposed models are able to interact with chemical spaces through natural language and symbolic notations, and have emerging extensions to incorporate multi-modal inputs.

Copied to clipboard

Zhenyu Lei, Zhen Tan, Song Wang, Yaochen Zhu, Zihan Chen, Yushun Dong, Jundong Li

DR-HM: Distill-then-Reinforce Training with Cognition-Aware Data Synthesis for Harmful Meme Detection (2026.findings-acl)

Challenge:	Recent studies suggest that the reasoning abilities of large language models (LLMs) grows with model size and pre-training data.
Approach:	They propose to combine quality filtering, conditional routing, and cooperative peer teaching to transfer knowledge from powerful teacher models to compact and transparent students.
Outcome:	Experiments show that QR-Distill is superior to traditional methods.

Copied to clipboard

Zihan Cheng, Jianxiang Ma, Xiaocui Yang, Peidong Wang, Wen Zhang, Shi Feng, Daling Wang, Yifei Zhang, Mingfu Zhang

X-Class: Text Classification with Extremely Weak Supervision (2021.naacl-main)

Challenge:	Current methods for harmful meme detection lack the knowledge required to identify such hate . current methods lack the ability to identify cultural stereotypes and visual metaphors .
Approach:	They propose a framework that decomposes meme analysis into a human-inspired reasoning process . they propose DR-HM to transfer knowledge from closed-source models while mitigating biases .
Outcome:	The proposed framework outperforms existing methods on three benchmark datasets.

Copied to clipboard

Zihan Wang, Dheeraj Mekala, Jingbo Shang

Goal-Driven Explainable Clustering via Language Descriptions (2023.emnlp-main)

Challenge:	Weak supervision is a problem in text classification, but it requires corpusspecific knowledge.
Approach:	They propose a framework for extremely weak supervision that can be used to train a text classifier.
Outcome:	The proposed framework outperforms seed-driven weakly supervised methods on 7 benchmark datasets.

Copied to clipboard

Zihan Wang, Jingbo Shang, Ruiqi Zhong

Explainability & Interpretability in NLP

Challenge:	Existing formulations neither consider the users’ goals nor explain clusters’ meanings.
Approach:	They propose a task formulation that represents both the goal and the explanations as free-form language descriptions.
Outcome:	The proposed method produces more accurate and goal-related explanations than previous methods.

Evaluating the Smooth Control of Attribute Intensity in Text Generation with LLMs (2024.findings-acl)

Copied to clipboard

Shang Zhou, Feng Yao, Chengyu Dong, Zihan Wang, Jingbo Shang

Multi-step Problem Solving Through a Verifier: An Empirical Analysis on Model-induced Process Supervision (2024.findings-emnlp)

Challenge:	Controllable text generation is increasingly tailored to individual preferences.
Approach:	They propose to evaluate the attribute intensity of text generated by large language models on five different attributes for error, variation of the generated sentence's intensities and relevance to the generation questions.
Outcome:	The proposed methods are based on Elo rating system and GPT4 and are able to be trained without training.

Copied to clipboard

Zihan Wang, Yunxuan Li, Yuexin Wu, Liangchen Luo, Le Hou, Hongkun Yu, Jingbo Shang

NaturalGAIA: A Verifiable Benchmark and Hierarchical Framework for Long-Horizon GUI Tasks (2026.acl-long)

Challenge:	a method for process supervision has shown significant improvements in multi-step problem solving . despite the advances in process supervision, there are still easily observable mistakes in state-of-the-art LLMs.
Approach:	They propose a method for automating data curation by using a trained verifier to evaluate intermediate steps generated by a reasoner.
Outcome:	The proposed method improves the performance of PaLM 2 on math and coding tasks.

Copied to clipboard

Zihan Zheng, Tianle Cui, Taoran Wang, Fengtao Wang, Jiahui Pan, Lewei He, Qianglong Chen

Challenge:	Current research faces an "Evaluation-Realism Dilemma" due to unstable MLLM judges or manual verification.
Approach:	They propose a verifiable evaluation dataset grounded in real-world human GUI intents.
Outcome:	The proposed framework outperforms the state-of-the-art framework in achieving a weighted pathway success rate of 45.6% while reducing token consumption and execution time by 76%.

PromptBERT: Improving BERT Sentence Embeddings with Prompts (2022.emnlp-main)

Copied to clipboard

Ting Jiang, Jian Jiao, Shaohan Huang, Zihan Zhang, Deqing Wang, Fuzhen Zhuang, Furu Wei, Haizhen Huang, Denvy Deng, Qi Zhang

Dual Slot Selector via Local Reliability Verification for Dialogue State Tracking (2021.acl-long)

Challenge:	Existing research shows that BERT and RoBERTa are poor in sentence embeddings due to static token embeddable bias and ineffective BERT layers.
Approach:	They propose a novel contrastive learning method for better sentence embeddings by using a template denoising technique.
Outcome:	The proposed method achieves 2.29 and 2.58 points of improvement compared to SimCSE and RoBERTa in the unsupervised setting.

Copied to clipboard

Jinyu Guo, Kai Shuang, Jijie Li, Zihan Wang

Enhancing Learning-Based Binary Code Similarity Detection Model through Adversarial Training with Multiple Function Variants (2024.findings-emnlp)

Challenge:	Existing approaches to predict dialogue state from scratch are inefficient and lead to errors . empirical results show that our method achieves 56.93%, 60.73%, and 58.04% joint accuracy on multi-domain conversations .
Approach:	They propose a dual-stage dialogue state tracking method that uses a slot selector and a Slot Value generator to predict the current dialogue state.
Outcome:	The proposed method achieves 56.93%, 60.73%, and 58.04% joint accuracy on multi-domain conversations.

Copied to clipboard

Lichen Jia, Chenggang Wu, Bowen Tang, Peihua Zhang, Zihan Jiang, Yang Yang, Ning Liu, Jingfeng Zhang, Zhe Wang

A Closer Look into Mixture-of-Experts in Large Language Models (2025.findings-naacl)

Challenge:	Existing Learning-Based Binary Code Similarity Detection (LB-BCSD) methods exhibit lower accuracy in recognizing functions with the same functionality but different implementations.
Approach:	They propose a gradient-guided adversarial attack method based on critical code called FuncFooler which perturbs critical code to generate multiple variants of the same function.
Outcome:	The proposed method increases the accuracy of the current Learning-Based Binary Code Similarity Detection (LB-BCSD) model by 5%-7%.

Copied to clipboard

Ka Man Lo, Zeyu Huang, Zihan Qiu, Zili Wang, Jie Fu

WeDef: Weakly Supervised Backdoor Defense for Text Classification (2022.emnlp-main)

Challenge:	Mixture-of-experts (MoE) architectures are gaining increasing attention for their unique properties and remarkable performance.
Approach:	They propose a mixture-of-experts architecture that allows for model scaling without sacrificing computational efficiency.
Outcome:	The proposed model increases model size without sacrificing computational efficiency . the proposed model is modular and can be used by a broad spectrum of practitioners .

Copied to clipboard

Lesheng Jin, Zihan Wang, Jingbo Shang

Less than One-shot: Named Entity Recognition via Extremely Weak Supervision (2023.findings-emnlp)

Challenge:	Existing backdoor defense methods are only effective for limited trigger types . Existing methods only work against certain types of triggers .
Approach:	They propose a weakly supervised backdoor defense framework WeDef to defend different trigger types at once . they define reliability of samples based on whether predictions agree with their labels .
Outcome:	The proposed framework outperforms existing backdoor defense methods against popular trigger-based attacks.

Copied to clipboard

Letian Peng, Zihan Wang, Jingbo Shang

Enhancing the General Agent Capabilities of Low-Paramter LLMs through Tuning and Multi-Branch Reasoning (2024.findings-naacl)

Challenge:	Named entity recognition (NER) problem is performed under extremely weak supervision . XWS setting is considered weaker than 1-shot since example entity is given in context-free way .
Approach:	They propose a method that uses extremely weak supervision to train named entity recognition models.
Outcome:	The proposed method outperforms the state-of-the-art few-shot methods with 1-shot supervision and ChatGPT annotations significantly.

Copied to clipboard

Qinhao Zhou, Zihan Zhang, Xiang Xiang, Ke Wang, Yuchuan Wu, Yongbin Li

Retrieve-Plan-Generation: An Iterative Planning and Answering Framework for Knowledge-Intensive LLM Generation (2024.emnlp-main)

Challenge:	Open-source pre-trained Large Language Models exhibit strong language understanding and generation capabilities, making them highly successful in a variety of tasks.
Approach:	They propose a method to construct agent-specific data using GPT-4 and supervised fine-tuning . they find that supervised tunning can significantly reduce hallucination outputs and formatting errors in agent tasks .
Outcome:	The proposed method improves on five agent tasks of AgentBench.

Copied to clipboard

Yuanjie Lyu, Zihan Niu, Zheyong Xie, Chao Zhang, Tong Xu, Yang Wang, Enhong Chen

WebCPM: Interactive Web Search for Chinese Long-form Question Answering (2023.acl-long)

Challenge:	Large language models (LLMs) often produce factual errors due to limited internal knowledge.
Approach:	They propose a retrieval-augmented generation framework that generates plan tokens to guide subsequent generation.
Outcome:	The proposed framework improves the accuracy of large language models with external knowledge sources.

Copied to clipboard

Yujia Qin, Zihan Cai, Dian Jin, Lan Yan, Shihao Liang, Kunlun Zhu, Yankai Lin, Xu Han, Ning Ding, Huadong Wang, Ruobing Xie, Fanchao Qi, Zhiyuan Liu, Maosong Sun, Jie Zhou

Commentary Generation from Data Records of Multiplayer Strategy Esports Game (2024.naacl-srw)

Challenge:	Long-form question answering requires two procedures: information retrieval and information synthesis.
Approach:	They propose a Chinese long-form question answering dataset called WebCPM . the dataset is based on a web search interface that engages with a search engine in real time .
Outcome:	The proposed dataset generates answers that are no worse than human-written ones . the dataset is the first Chinese LFQA dataset .

Copied to clipboard

Zihan Wang, Naoki Yoshinaga

From Cross-Task Examples to In-Task Prompts: A Graph-Based Pseudo-Labeling Framework for In-context Learning (2025.findings-emnlp)

Challenge:	Esports play logs are expensive for human experts to provide individual games with play-by-play commentaries.
Approach:	They first build large-scale esports data-to-text datasets that pair structured data and commentaries from a popular eSports game, League of Legends.
Outcome:	The proposed model can generate game commentaries from esports’ data records while examining the impact of the pre-trained language models.

Copied to clipboard

Zihan Chen, Song Wang, Xingbo Fu, Chengshuai Shi, Zhenyu Lei, Cong Shen, Jundong Li

CrossWeigh: Training Named Entity Tagger from Imperfect Annotations (D19-1)

Challenge:	In-context learning (ICL) enables large language models to perform novel tasks without parameter updates by conditioning on a few input-output examples.
Approach:	They propose a cost-efficient two-stage pipeline that reduces reliance on LLMs for data labeling.
Outcome:	The proposed pipeline reduces reliance on LLMs for data labeling . it leverages readily available cross-task examples to prompt an LLM and pseudo-label a small set of target task instances.

Copied to clipboard

Zihan Wang, Jingbo Shang, Liyuan Liu, Lihao Lu, Jiacheng Liu, Jiawei Han

CodeContests+: High-Quality Test Case Generation for Competitive Programming (2025.findings-emnlp)

Challenge:	Named entity recognition (NER) models can identify labels in 5.38% of test sentences . a framework to handle label mistakes during NER model training is proposed .
Approach:	They propose a framework to manually correct label mistakes in named entity recognition (NER) they aim to improve the accuracy of models by re-evaluating popular models on corrected test sets .
Outcome:	The proposed framework can detect label mistakes in 5.38% of test sentences . the proposed framework improves on three datasets with a high-performance model .

Copied to clipboard

Zihan Wang, Siyao Liu, Yang Sun, Ming Ding, Hongyan Li

Extending Multilingual BERT to Low-Resource Languages (2020.findings-emnlp)

Challenge:	Competitive programming has become a key task for training and evaluating large language models . but test cases of competitive programming problems are often difficult to obtain .
Approach:	They propose an LLM-based agent system that creates high-quality test cases for competitive programming problems.
Outcome:	The proposed system improves code tests on a CodeContests dataset with pass/fail labels.

Copied to clipboard

Zihan Wang, Karthikeyan K, Stephen Mayhew, Dan Roth

SUT: Active Defects Probing for Transcompiler Models (2023.emnlp-main)

Challenge:	Multilingual BERT (M-BERT) has been a huge success in both supervised and zero-shot cross-lingual transfer learning.
Approach:	They propose a simple but effective approach to extend multilingual BERT to any new language and show an increase in F1 on M-BERT and new languages.
Outcome:	The proposed approach improves on languages already in M-BERT and out of it on other languages.

Copied to clipboard

Mengnan Qi, Yufan Huang, Maoquan Wang, Yongqiang Yao, Zihan Liu, Bin Gu, Colin Clement, Neel Sundaresan

Machine Translation

Challenge:	Existing datasets are often criticized for their lack of granularity, which can mask deficiencies in basic syntactic elements that humans care about.
Approach:	They propose a new program translation metrics that address basic syntax errors . they propose BLUE, CodeBLUE and computation accuracy metrics which address these errors based on a highly interpretable evaluation harness.
Outcome:	The proposed model passes the unit tests with a 26.15% pass rate compared to previous models .

Multi-Defendant Legal Judgment Prediction via Hierarchical Reasoning (2023.findings-emnlp)

Copied to clipboard

Yougang Lyu, Jitai Hao, Zihan Wang, Kai Zhao, Shen Gao, Pengjie Ren, Zhumin Chen, Fang Wang, Zhaochun Ren

Identifying Collective Intelligence Factor in LLM Agent Groups for Generalizable Multi-Agent System Design (2026.findings-acl)

Challenge:	Existing methods for predicting judgment results for multiple defendants are ineffective.
Approach:	They propose a method to predict the judgment results for each defendant in multi-defendant cases . they formalize the multi-diffendant judgment process as hierarchical reasoning chains .
Outcome:	The proposed method can predict the judgment results for multiple defendants in multi-defendant cases.

Copied to clipboard

Zhilun Zhou, Zihan Liu, Jiahe Liu, Yihan Wang, Qingyu Shao, Fengli Xu, Depeng Jin, Yong Li

Challenge:	Prior studies have focused on designing customized MAS for specific tasks . a critical research question remains: do LLM agent groups exhibit a form of "general intelligence"
Approach:	They find a Collective Intelligence factor in human groups that captures their general capability.
Outcome:	The proposed model predicts the ACI factor based on the features of LLM agent groups and can improve generalization abilities.

LLaSE-G1: Incentivizing Generalization Capability for LLaMA-based Speech Enhancement (2025.acl-long)

Copied to clipboard

Boyi Kang, Xinfa Zhu, Zihan Zhang, Zhen Ye, Mingshuai Liu, Ziqian Wang, Yike Zhu, Guobin Ma, Jun Chen, Longshuai Xiao, Chao Weng, Wei Xue, Lei Xie

Generalizing Few-Shot Named Entity Recognizers to Unseen Domains with Type-Related Features (2023.findings-emnlp)

Challenge:	Recent advances in language models have demonstrated strong capabilities in semantic understanding and contextual modeling.
Approach:	They propose a LLaMA-based language model that incentivizes generalization capabilities for speech enhancement.
Outcome:	The proposed language model outperforms prior task-specific discriminative and generative models in acoustic enhancement tasks.

Copied to clipboard

Zihan Wang, Ziqi Zhao, Zhumin Chen, Pengjie Ren, Maarten de Rijke, Zhaochun Ren

CoRAG: Enhancing Hybrid Retrieval-Augmented Generation through a Cooperative Retriever Architecture (2025.findings-emnlp)

Challenge:	Few-shot named entity recognition methods struggle with out-of-domain (OOD) examples due to their reliance on manual labeling for the target domain.
Approach:	They propose a framework to enable generalization to an unseen target domain with only a few labeled examples.
Outcome:	The proposed framework achieves significant performance improvements on in-domain and cross-domain datasets.

Copied to clipboard

Zaiyi Zheng, Song Wang, Zihan Chen, Yaochen Zhu, Yinhan He, Liangjie Hong, Qi Guo, Jundong Li

From 128K to 4M: Efficient Training of Ultra-Long Context Large Language Models (2026.findings-acl)

Challenge:	Existing methods only retrieve related documents from local neighbors or subgraphs in the knowledge base, which often miss relevant information located further away from a global view.
Approach:	Hybrid-RAG combines textual documents and graph-structured relational information for RAG . existing methods only retrieve related documents from local neighbors or subgraphs in the knowledge base .
Outcome:	Hybrid-RAG combines textual documents and graph-structured relational information . existing methods only retrieve related documents from local neighbors or subgraphs in the knowledge base .

Copied to clipboard

Chejian Xu, Wei Ping, Peng Xu, Zihan Liu, Boxin Wang, Mohammad Shoeybi, Bo Li, Bryan Catanzaro

A Benchmark on Extremely Weakly Supervised Text Classification: Reconcile Seed Matching and Prompting Approaches (2023.findings-acl)

Challenge:	Long-context capabilities are essential for document and video understanding, in-contact learning, and inference-time scaling.
Approach:	They propose an efficient training recipe for building ultra-long context LLMs from aligned instruct model, pushing the boundaries of context lengths from 128K to 1M, 2M, and 4M tokens.
Outcome:	The proposed model extends the context window while maintaining short context capabilities while maintaining the performance of the existing model.

Copied to clipboard

Zihan Wang, Tianle Wang, Dheeraj Mekala, Jingbo Shang

ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversation (2023.findings-emnlp)

Challenge:	Existing methods for XWS-TC rely on minimal human guidance . X-WS-tc methods require no humanannotated datasets .
Approach:	They propose a benchmarking method to compare two approaches to XWS-TC . they use seed-matching and prompting a language model with instructions to decode label words .
Outcome:	The proposed methods are more tolerant to human guidance and more robust to model-based methods.

Copied to clipboard

Zi Lin, Zihan Wang, Yongqi Tong, Yangkun Wang, Yuxin Guo, Yujia Wang, Jingbo Shang

Incorporating Hierarchy into Text Encoder: a Contrastive Learning Approach for Hierarchical Text Classification (2022.acl-long)

Challenge:	toxicity detection has been largely based on social media content, leaving the unique challenges inherent to real-world user-AI interactions insufficiently explored.
Approach:	They propose a benchmark to detect toxicity in real-world user-AI conversations . they compare existing models with social media content to find toxicity .
Outcome:	The proposed benchmark reveals that existing models fail to recognize toxicity in real-world user-AI conversations.

Copied to clipboard

Zihan Wang, Peiyi Wang, Lianzhe Huang, Xin Sun, Houfeng Wang

Shall We Pretrain Autoregressive Language Models with Retrieval? A Comprehensive Study (2023.emnlp-main)

Challenge:	Existing methods encode text and label hierarchy separately and mix their representations for classification, where the hierarchy remains unchanged for all input text.
Approach:	They propose to embed hierarchy into a text encoder by combining input and output data to generate a hierarchy-aware representation.
Outcome:	Extensive experiments on three benchmark datasets verify the effectiveness of the proposed model.

Copied to clipboard

Boxin Wang, Wei Ping, Peng Xu, Lawrence McAfee, Zihan Liu, Mohammad Shoeybi, Yi Dong, Oleksii Kuchaiev, Bo Li, Chaowei Xiao, Anima Anandkumar, Bryan Catanzaro

InstructPart: Task-Oriented Part Segmentation with Instruction Reasoning (2025.acl-long)

Challenge:	a recent study shows that retrieval-augmented LMs can improve text generation quality and accuracy.
Approach:	They propose a model that reproduces RETRO parameters while retrieving a text corpus . they find RETRO outperforms GPT on text generation with less repetition .
Outcome:	The proposed model outperforms standard retrieval-augmented GPT and retrieval augmented GTP on text generation and accuracy tasks.

Copied to clipboard

Zifu Wan, Yaqi Xie, Ce Zhang, Zhiqiu Lin, Zihan Wang, Simon Stepputtis, Deva Ramanan, Katia P. Sycara

AssistSR: Task-oriented Video Segment Retrieval for Personal AI Assistant (2022.findings-emnlp)

Challenge:	Large multimodal foundation models perceive objects as indivisible, overlooking the components that constitute them.
Approach:	They propose a novel benchmark for large multimodal foundation models comprising hand-labeled part segmentation annotations and task-oriented instructions to evaluate their performance.
Outcome:	The proposed benchmark improves performance of current models in understanding and executing part-level tasks within everyday contexts.

Copied to clipboard

Weixian Lei, Difei Gao, Yuxuan Wang, Dongxing Mao, Zihan Liang, Lingmin Ran, Mike Zheng Shou

Beyond the Granularity: Multi-Perspective Dialogue Collaborative Selection for Dialogue State Tracking (2022.acl-long)

Challenge:	Currently, personal AI assistants on the phone and AR glasses can assist our daily life in addressing our questions like "how to adjust the date for this watch?"
Approach:	They propose a task that asks a question about affordance of items in our daily life . they construct a dataset that contains 3.2k multimodal questions on 1.6k video segments .
Outcome:	The proposed task outperforms baseline methods while still having room for improvement in the future.

Copied to clipboard

Jinyu Guo, Kai Shuang, Jijie Li, Zihan Wang, Yixuan Liu

AnyMAC: Cascading Flexible Multi-Agent Collaboration via Next-Agent Prediction (2025.emnlp-main)

Challenge:	Experimental results show that task-oriented dialogue systems have attracted growing attention and achieved substantial progress.
Approach:	They propose a method that dynamically selects relevant dialogue contents for each slot . they retrieve turn-level utterances and evaluate their relevance to the slot from three perspectives .
Outcome:	The proposed method achieves state-of-the-art performance on MultiWOZ 2.1 and MultiWOz 2.2 and superior performance on multiple mainstream benchmark datasets.

Copied to clipboard

Song Wang, Zhen Tan, Zihan Chen, Shuang Zhou, Tianlong Chen, Jundong Li

CIRAG: Construction–Integration Retrieval and Adaptive Generation for Multi-hop Question Answering (2026.acl-long)

Challenge:	Existing methods for multi-agent collaboration rely on static or graph-based topologies lacking flexibility and adaptability.
Approach:	They propose a new framework that rethinks multi-agent coordination through a sequential structure rather than a graph structure.
Outcome:	The proposed method achieves superior performance while significantly reducing communication overhead.

Copied to clipboard

Zili Wei, Yilin Wang, Xiaocui Yang, Shi Feng, Weidong Bao, Daling Wang, Zihan Wang, Yifei Zhang

SpiderFlow: Efficient Topology-Aware Scheduling for LLM Training Across Decentralized GPU Clusters (2026.acl-long)

Challenge:	Existing methods for iterative retrieval-augmented generation (iRAG) suffer from greedy single-path expansion and granularity–demand mismatch .
Approach:	They propose a model that constructs candidate triples and history-conditionally integrates them to distill core triples to generate the next-hop query.
Outcome:	The proposed model mitigates the greedy single-path expansion and granularity–demand mismatch by preserving multiple plausible evidence chains.

Copied to clipboard

Zihan Chang, Shuibing He, Bo Zhou, Sheng Xiao, Siling Yang, Rui Wang, Zhe Pan

Learning Adaptive Axis Attentions in Fine-tuning: Beyond Fixed Sparse Attention Patterns (2022.findings-acl)

Challenge:	Existing approaches to training large language models lack topologyaware task scheduling mechanisms and model parallelization strategies.
Approach:	They propose a topology-aware scheduling system specifically designed for decentralized GPU clusters . they propose heuristic methods at the inter-cluster level with ILP-based optimization within clusters.
Outcome:	The proposed system reduces job completion time by 1.2-1.3 and improves throughput by 1.12-1.25 . it also reduces scheduling overhead by 20-90 on average compared to state-of-the-art scheduling systems.

Copied to clipboard

Zihan Wang, Jiuxiang Gu, Jason Kuen, Handong Zhao, Vlad Morariu, Ruiyi Zhang, Ani Nenkova, Tong Sun, Jingbo Shang

InfoGain-RAG: Boosting Retrieval-Augmented Generation through Document Information Gain-based Reranking and Filtering (2025.emnlp-main)

Challenge:	Adaptive Axis Attention learns different attention patterns for each task and model layer . sparse attention patterns do not improve the run time of the models but they reduce model memory requirements .
Approach:	They propose a method that learns different attention patterns for each Transformer layer . they propose 'adaptive axis attention' method that identifies important tokens .
Outcome:	The proposed method does not require pre-training to accommodate sparse attention patterns.

Copied to clipboard

Zihan Wang, Zihan Liang, Zhou Shao, Yufei Ma, Huangyu Dai, Ben Chen, Lingtao Mao, Chenyi Lei, Yuqing Ding, Han Li

MessToClean: Evidence-Grounded Structure-Preserving Reconstruction for Real-World Degraded Exam Paper Images (2026.acl-long)

Challenge:	Retrieval-Augmented Generation (RAG) frameworks struggle with identifying whether retrieved documents meaningfully contribute to answer generation.
Approach:	They propose a document-related metric to quantify the contribution of retrieved documents to correct answer generation.
Outcome:	The proposed framework outperforms existing approaches on both single and multiple retrieval paradigms.

Copied to clipboard

Jiayi Tuo, Cheng Tang, Zihan Wang, Chenyue Zhou, Yao Li, Yanbiao Ma, Chao Wang, Wei Dai, Mingxuan Wang, Shitong Qin, Ziwei Zhao

EviReport: From Reasoned Outlines to Evidence Tracked Long-Form Reports (2026.findings-acl)

Challenge:	Existing Multimodal Large Language Models (MLLMs) fail under RDEI, leading to disrupted structure and evidence-unsupported hallucinations.
Approach:	They propose a backbone-agnostic, evidence-driven pipeline that treats off-the-shelf MLLMs as interchangeable components to improve stem consistency and figure consistency.
Outcome:	The proposed pipeline improves stem consistency by 1.01-3.18%, figure consistency by 0.50-49.16%, and refusal F1 by 1.06-10.88% across question types.

Copied to clipboard

Zihan Liu, Jianhui li, Zexin Wang, Fei Sun, Jingjing LI, Zheyuan Li, Ke Xiang, Hang Cui, Houhua Gong, Changhua Pei, Gaogang Xie

“Average” Approximates “First Principal Component”? An Empirical Analysis on Representations from Neural Language Models (2021.emnlp-main)

Challenge:	Evidence-intensive reports often produce fluent but under-supported drafts . eviReport is an evidence-grounded workflow for automated long-form report generation .
Approach:	They propose an evidence-tracked workflow that organizes corpus evidence into compact, traceable units and retrieves query-relevant subgraphs into retrieval-ready packages.
Outcome:	The proposed workflow outperforms baselines in factual coverage, factual accuracy and visual evidence integration.

Copied to clipboard

Zihan Wang, Chengyu Dong, Jingbo Shang

Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models (2024.emnlp-main)

Challenge:	Contextualized representations have been used in various NLP tasks, but their nature remains a mystery.
Approach:	They propose to use a property to estimate the power of contextualized representations . they show that the average representation shares almost the same direction as the first principal component .
Outcome:	The proposed representations share the same direction as the first principal component . the results suggest that the property is intrinsic to the distribution of representations .

Copied to clipboard

Zihan Wang, Deli Chen, Damai Dai, Runxin Xu, Zhuoshu Li, Yu Wu

MTRouter: Cost-Aware Multi-Turn LLM Routing with History–Model Joint Embeddings (2026.acl-long)

Challenge:	Existing studies on parameter-efficient fine-tuning (PEFT) for dense-architecture LLMs are lacking.
Approach:	They propose an expert-specialized fine-tuning method that tunes the experts most relevant to downstream tasks while freezing the other experts.
Outcome:	The proposed method matches or surpasses full-parameter fine-tuning.

Copied to clipboard

Yiqun Zhang, Hao Li, Zihan Wang, Shi Feng, Xiaocui Yang, Daling Wang, Bo Zhang, Lei Bai, Shuyue Hu

Decompose, Prioritize, and Eliminate: Dynamically Integrating Diverse Representations for Multimodal Named Entity Recognition (2024.lrec-main)

Challenge:	Multi-turn, long-horizon tasks require dozens of sequential model calls per episode.
Approach:	They propose a cost-aware multi-turn LLM routing tool which encodes interaction history and candidate models into joint history–model embeddings and learns an outcome estimator from logged trajectories to predict turn-level model utility.
Outcome:	The proposed model reduces cost and performance by 58.7% on ScienceWorld and on Humanity’s Last Exam (HLE) and even reduces costs for held-out tasks.

Copied to clipboard

Zihao Zheng, Zihan Zhang, Zexin Wang, Ruiji Fu, Ming Liu, Zhongyuan Wang, Bing Qin

Answer is All You Need: Instruction-following Text Embedding via Answering the Question (2024.acl-long)

Challenge:	Existing research on multi-modal Named Entity Recognition (MNER) does not integrate all multi-modal representations to provide rich contextual information to improve NER.
Approach:	They propose an iterative reasoning framework that integrates all the diverse multi-modal representations following the strategy of "decompose, prioritize, and eliminate" . they propose to use hierarchically connected fusion layers to prioritize transitions from "easy-to-hard" and "coarse-to fine"
Outcome:	The proposed framework integrates all the diverse multi-modal representations following the strategy of "decompose, prioritize, and eliminate".

Copied to clipboard

Letian Peng, Yuwei Zhang, Zilong Wang, Jayanth Srinivasa, Gaowen Liu, Zihan Wang, Jingbo Shang

GitChameleon 2.0: Evaluating AI Code Generation Against Python Library Version Incompatibilities (2026.acl-long)

Challenge:	Existing methods for encoding instruction information fail to be sensitive to clearer criteria like “evaluate similarity based on emotion” . instead, we propose a different approach, which treats the instruction as a “question” about the input text and encodes the expected answers to obtain the representation accordingly.
Approach:	They propose a text embedder that captures characteristics of texts specified by user instructions clarifying the similarity criterion.
Outcome:	The proposed model improves instruction-following capabilities when applied to large language models and encoder-based LMs.

Copied to clipboard

Diganta Misra, Nizar Islah, Victor May, Brice Rauby, Zihan Wang, Justine Gehring, Antonio Orvieto, Muawiz Sajjad Chaudhary, Eilif B. Muller, Irina Rish, Samira Ebrahimi Kahou, Massimo Caccia

Programming Languages in NLP

Challenge:	Existing code evolution benchmarks lack execution-based evaluation for generating code compliant with specific library versions.
Approach:	They propose a new Python code completion problem that evaluates the ability of large language models to perform version-conditioned code generation.
Outcome:	The proposed benchmarks show that state-of-the-art systems can perform version-conditioned code generation with high success rates.

LLM×MapReduce: Simplified Long-Sequence Processing using Large Language Models (2025.acl-long)

Copied to clipboard

Zihan Zhou, Chong Li, Xinyi Chen, Shuo Wang, Yu Chao, Zhili Li, Haoyu Wang, Qi Shi, Zhixing Tan, Xu Han, Xiaodong Shi, Zhiyuan Liu, Maosong Sun