Papers by Hai Wu
Towards Enhanced Immersion and Agency for LLM-based Interactive Drama (2025.acl-long)
Copied to clipboard
| Challenge: | Existing studies have focused on the role of immersion and agency in interactive drama. |
| Approach: | They propose a playwriting-guided generation method that helps LLMs craft dramatic stories with substantially improved structures and narrative quality. |
| Outcome: | The proposed method improves storytelling quality and immersion and agency, while allowing agents to refine their reactions to align with the player’s intentions. |
GMSA: Enhancing Context Compression via Group Merging and Layer Semantic Alignment (2026.acl-long)
Copied to clipboard
Jiwei Tang, Zhicheng Zhang, Shunlong Wu, Jingheng Ye, Lichen Bai, Zitai Wang, Tingwei Lu, Lin Hai, Yiming Zhao, Hai-Tao Zheng, Hong-Gee Kim
| Challenge: | Large Language Models (LLMs) have achieved remarkable performance across NLP tasks . however, in long-context scenarios, they face high computational cost and information redundancy. |
| Approach: | They propose an encoder-decoder context compression framework that generates a compact sequence of soft tokens for downstream tasks. |
| Outcome: | Experiments show that GMSA outperforms baselines on multiple long-context question answering and summarization benchmarks while maintaining low end-to-end latency. |
X-TURING: Towards an Enhanced and Efficient Turing Test for Long-Term Dialogue Agents (2025.acl-long)
Copied to clipboard
| Challenge: | Traditional Turing test limits each participant to one message at a time and requires constant human participation. |
| Approach: | They propose to enhance the original Turing test with a burst dialogue pattern, allowing more dynamic exchanges using consecutive messages. |
| Outcome: | The proposed test improves the original test with a burst dialogue pattern, allowing more dynamic exchanges using consecutive messages. |
Forging Multiple Training Objectives for Pre-trained Language Models via Meta-Learning (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Empirical studies show that learning multiple training objectives in a single model makes the learned language representation barely converge to the desired optimum. |
| Approach: | They propose a meta-learning-based adaptive sampler which learns latent sampling pattern on arbitrary pre-training objectives. |
| Outcome: | Empirical studies show that learning multiple objectives in a single model makes it difficult to achieve the desired optimum. |
Extrapolating Multilingual Understanding Models as Multilingual Generators (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing multilingual understanding models are not capable of generating high-quality text compared with decoder-based causal language models. |
| Approach: | They propose a method to adapt a multilingual encoder to a language generator with a small number of additional parameters. |
| Outcome: | The proposed approach outperforms initialization-based methods with 9.4 BLEU on machine translation, 8.1 Rouge-L on question generation, and 5.5 METEOR on story generation. |
Gradient-Boosted Decision Tree for Listwise Context Model in Multimodal Review Helpfulness Prediction (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing studies have shown that FCNNs perform inefficient splitting for review features, making it difficult to clearly differentiate helpful from unhelpful reviews. |
| Approach: | They propose a listwise attention network that captures the MRHP ranking context and a pairwise optimization objective that enhances model generalization. |
| Outcome: | The proposed framework achieves state-of-the-art results and polished generalization performance on two large-scale MRHP benchmark datasets. |
Adaptive Contrastive Learning on Multimodal Transformer for Review Helpfulness Prediction (2022.emnlp-main)
Copied to clipboard
| Challenge: | Modern review helpfulness prediction systems focus on polishing cross-modal representations and suffer from inferior optimization. |
| Approach: | They propose a method to polish cross-modal relation representations by learning mutual information through contrastive learning. |
| Outcome: | The proposed framework outperforms baselines and achieves state-of-the-art results on two publicly available datasets. |
ToM: Leveraging Tree-oriented MapReduce for Long-Context Reasoning in Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Experimental results show ToM outperforms existing divide-and-conquer frameworks . RAG relies on similarity-based rankings to retrieve and reason over chunks based on logical coherence . |
| Approach: | They propose a Tree-oriented MapReduce framework for long-context reasoning . it leverages the hierarchical structure of long documents by constructing a DocTree . |
| Outcome: | Experimental results show that ToM outperforms existing divide-and-conquer frameworks and RAGs . the proposed framework improves logical coherence and long-context reasoning on 70B+ LLMs compared to existing approaches . |
Evolving Chinese Spelling Correction with Corrector-Verifier Collaboration (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent methods address Chinese Spelling Correction (CSC) with either BERT-based models or large language models (LLMs) however, both of them face challenges. |
| Approach: | They propose a model collaboration pipeline to iteratively optimize a BERT-based corrector. |
| Outcome: | The proposed model outperforms existing methods and outperformed human annotation methods. |
Agent-Pro: Learning to Evolve via Policy-Level Reflection and Optimization (2024.acl-long)
Copied to clipboard
Wenqi Zhang, Ke Tang, Hai Wu, Mengna Wang, Yongliang Shen, Guiyang Hou, Zeqi Tan, Peng Li, Yueting Zhuang, Weiming Lu
| Challenge: | Large Language Models (LLMs) are designed as specific task solvers with sophisticated prompt engineering, but are inherently incapacitating to address complex dynamic scenarios. |
| Approach: | They propose an LLM-based agent with policy-level reflection and optimization that can learn from interactive experiences and progressively elevate its behavioral policy. |
| Outcome: | The proposed agent outperforms vanilla LLM and specialized models in blackjack and Texas hold’em. |
Sentence Representation Learning with Generative Objective rather than Contrastive Objective (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing sentences-level training objectives focus on acquiring sentence-level representations, but they lack effective self-supervised objectives. |
| Approach: | They propose a generative self-supervised learning objective based on phrase reconstruction to improve sentence representation. |
| Outcome: | Empirical results show that the proposed objective outperforms current methods on STS benchmarks and retrieval and reranking tasks. |
Rethinking Masked Language Modeling for Chinese Spelling Correction (2023.acl-long)
Copied to clipboard
| Challenge: | Existing CSC models over-fit the error model while under-fitting the language model, resulting in poor generalization to out-of-distribution error patterns. |
| Approach: | They propose to use a multi-domain benchmark LEMON to assess the open domain generalization of Chinese Spelling Correction models. |
| Outcome: | The proposed method achieves state-of-the-art results on SIGHAN, ECSpell, and LEMON. |
Attack Named Entity Recognition by Entity Boundary Interference (2024.lrec-main)
Copied to clipboard
| Challenge: | Named Entity Recognition (NER) is a cornerstone natural language processing task . despite its robustness, studies on its robustity are lacking. |
| Approach: | They propose a one-word modification NER attack that strategically inserts a new boundary into the sentence and triggers the model to make a wrong recognition. |
| Outcome: | The proposed method is effective on English and Chinese models with 70%-90% success rate. |
Unveiling Vulnerability of Self-Attention (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing studies focus on manipulating word inputs, but they lack generalization to versatile real-world attacks. |
| Approach: | They propose a powerful perturbation technique which perturbs the attention scores within the SA matrices via meticulously crafted attention masks. |
| Outcome: | The proposed perturbation technique achieves high attack success rate (98%) and low cost. |
Chinese Spelling Corrector Is Just a Language Learner (2024.findings-acl)
Copied to clipboard
| Challenge: | a recent study shows that self-supervised learning can improve Chinese spelling correction by removing errors from training data. |
| Approach: | They propose a method that decodes Chinese spelling correction models using noise . they say it outperforms current methods that rely on annotated errors . |
| Outcome: | The proposed method outperforms the confusion set in specific domains because there are no errors in the training data. |
Sign2Vis: Automated Data Visualization from Sign Language (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to translate natural language descriptions into visualization queries focus on spoken languages, not sign languages. |
| Approach: | They propose a sign language interface that enables the DHH community to engage more fully with data analysis. |
| Outcome: | The proposed interface can be used by the deaf and hard-of-hearing community. |
Code Summarization with Structure-induced Transformer (2021.findings-acl)
Copied to clipboard
| Challenge: | Code summarization (CS) is a promising area in recent language understanding . previous work using structurebased traversal or non-sequential models to learn structural program semantics has shown no performance gain . |
| Approach: | They propose to use a structure-based traversal model to learn structural program semantics to generate human language automatically for programming language in the format of source code. |
| Outcome: | Experiments show that the proposed method achieves state-of-the-art on benchmarks. |
From Role-Play to Drama-Interaction: An LLM Solution (2024.findings-acl)
Copied to clipboard
| Challenge: | aristotle defined drama as a form of storytelling that involves a predefined storyline, emotions and thoughts. |
| Approach: | They propose to use LLMs to create an immersive mode of storytelling . they propose to create a backbone drama LLM to drive the playing process . |
| Outcome: | The proposed model can be used to drive the playing process, the authors say . it can be compared with existing models and can be evaluated on multiple scenarios. |
Sentence-aware Contrastive Learning for Open-Domain Passage Retrieval (2022.acl-long)
Copied to clipboard
| Challenge: | Existing studies focus on improving negative sampling strategy or extra pretraining for dense passage representations, but these studies are not capturing passage with internal representation conflicts. |
| Approach: | They propose a model with a smaller granularity to capture internal representation conflicts . they introduce a negative sampling strategy to encourage a diverse generation of sentence representations within the same passage. |
| Outcome: | The proposed model can be trained on three benchmark datasets to alleviate internal representation conflicts. |
Driving Chinese Spelling Correction from a Fine-Grained Perspective (2025.coling-main)
Copied to clipboard
| Challenge: | Existing evaluations for Chinese spelling correction lack nuanced typology for spelling errors, creating an "invisible" bottleneck . |
| Approach: | They propose a fine-grained evaluation principle for Chinese spelling correction (CSC) they categorize spelling errors into six different types and use it to evaluate models . |
| Outcome: | The proposed evaluation principle can be leveraged to enhance CSC training models. |
Mitigating Misleading Chain-of-Thought Reasoning with Selective Filtering (2024.lrec-main)
Copied to clipboard
| Challenge: | Large language models have demonstrated remarkable capabilities by leveraging chain-of-thought reasoning techniques to solve complex questions. |
| Approach: | They propose a method that assesses the entailment relationship between the question and the candidate reasoning chain and uses it to predict the answer. |
| Outcome: | The proposed approach improves the fine-tuned T5 baseline over the ScienceQA, ECQA, and LastLetter tasks. |
Instruction-Driven Game Engine: A Poker Case Study (2024.emnlp-demo)
Copied to clipboard
| Challenge: | a new concept allows users to create games by natural language instructions . the concept is based on the text-based game states, which are rendered to visual display . |
| Approach: | They propose an instruction-driven game engine that allows users to create games by natural language instructions. |
| Outcome: | The proposed concept allows users to create games simply by natural language instructions . initial progress lies in developing an IDGE for poker, which supports a wide range of poker variants . |
Semantic-Preserving Adversarial Code Comprehension (2022.coling-1)
Copied to clipboard
| Challenge: | Existing studies on improving PrLMs for source code comprehension have not found a way to improve both sides of the trade-off between the two aspects. |
| Approach: | They propose to use semantic-preserving code embeddings to find worst-case attacks while forcing the model to predict the correct labels under these worst cases. |
| Outcome: | The proposed model can stay robust against state-of-the-art attacks while boosting the performance of PrLMs for code. |
Open-Theatre: An Open-Source Toolkit for LLM-based Interactive Drama (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing tools for creating, modifying, and experimenting with interactive dramas are limited. |
| Approach: | They propose an open-source toolkit for creating configurable LLM-based interactive drama. |
| Outcome: | The proposed toolkit enhances narrative coherence and realistic behavior in interactions with agents. |
Game Development as Human-LLM Interaction (2025.acl-long)
Copied to clipboard
| Challenge: | Currently, game development is a highly specialized task that relies on a complex game engine powered by complex programming languages, preventing many gaming enthusiasts from handling it. |
| Approach: | They propose a chat game engine powered by LLM that allows everyone to develop a custom game using natural language through Human-LLM interaction. |
| Outcome: | The proposed engine is designed to support the development of custom games using natural language through Human-LLM interaction. |
Unfolding the Headline: Iterative Self-Questioning for News Retrieval and Timeline Summarization (2025.findings-naacl)
Copied to clipboard
| Challenge: | a new approach to timeline summarization is proposed for open-domain news content . large language models (LLMs) can be used to extract and organize news events from multiple documents . |
| Approach: | They propose a method to integrate Large Language Models into news timeline summarization by iterating on how events are linked and posing new questions. |
| Outcome: | The proposed system is able to generate and refresh chronological summaries based on documents retrieved in each round. |
Empower Nested Boolean Logic via Self-Supervised Curriculum Learning (2023.emnlp-main)
Copied to clipboard
| Challenge: | a new self-supervised learning method allows language models to generalize to much harder and longer-hop logic. |
| Approach: | They propose a self-supervised learning method to empower language models with boolean logic . they augment training data with nested booles and program training from simpler to harder ones . |
| Outcome: | The proposed method allows language models to generalize to much harder and longer-hop logic, which can hardly be learned through naive training. |