Papers by Xiaobao Wu
Full-Step-DPO: Self-Supervised Preference Optimization with Step-wise Rewards for Mathematical Reasoning (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to improve long-chain mathematical reasoning focus on the first erroneous step, but ignore all other steps and rely heavily on external signals. |
| Approach: | They propose a DPO framework that leverages step-wise rewards from the entire reasoning chain instead of optimizing only the first erroneous step. |
| Outcome: | The proposed framework improves on in-domain and out-of-domain mathematical reasoning benchmarks. |
Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level Embeddings (2025.findings-acl)
Copied to clipboard
Yubo Ma, Jinsong Li, Yuhang Zang, Xiaobao Wu, Xiaoyi Dong, Pan Zhang, Yuhang Cao, Haodong Duan, Jiaqi Wang, Yixin Cao, Aixin Sun
| Challenge: | Visualized Document Retrieval (VDR) uses large vision-language models to encode document pages into embeddings. |
| Approach: | They evaluate methods to reduce patch embeddings per page while minimizing performance degradation. |
| Outcome: | The proposed method maintains 98.2% of retrieval performance with only 11.8% of original memory usage and preserves 94.6% effectiveness at 2% memory footprint. |
Understanding and Preventing Entropy Collapse in RLVR with On-Policy Entropy Flow Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing RLVR algorithms suffer from entropy collapse, leading to premature determinism and unstable optimization. |
| Approach: | They propose an adaptive entropy flow balancing mechanism that rescales entropic-increasing and enotro-decreazing updates according to their contributions to enthroy change. |
| Outcome: | The proposed method outperforms existing RLVR algorithms on six reasoning benchmarks. |
Unsupervised Hallucination Detection by Inspecting Reasoning Processes (2025.emnlp-main)
Copied to clipboard
| Challenge: | Unsupervised hallucination detection aims to identify hallucines generated by large language models without relying on labeled data. |
| Approach: | They propose an unsupervised method to detect hallucinated content by large language models . they use internal representations intrinsic to factual correctness to prompt the model to verify the truthfulness of a given statement . |
| Outcome: | The proposed framework outperforms existing unsupervised methods and is fully unsupervised and low cost. |
Encoding and Controlling Global Semantics for Long-form Video Question Answering (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to find answers for long videos fail to reason over the whole sequence of video, leading to sub-optimal performance. |
| Approach: | They propose a state space layer to integrate global semantics into video . they use a gating unit to enable controllability over the flow of global semantic into visual representations. |
| Outcome: | The proposed framework is able to integrate global semantics into visual representations. |
Fact-Checking Complex Claims with Program-Guided Reasoning (2023.acl-long)
Copied to clipboard
| Challenge: | Fact-checking real-world claims often requires collecting multiple pieces of evidence and complex multi-step reasoning. |
| Approach: | They propose a novel fact-checking model that decomposes complex claims into simpler sub-tasks that can be solved using a shared library of specialized functions. |
| Outcome: | The proposed model outperforms seven baselines on two fact-checking datasets and has explicit output programs that benefit human debugging. |
Mitigating Data Sparsity for Short Text Topic Modeling by Topic-Semantic Contrastive Learning (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing topic models do not make full use of word co-occurrence information to model latent topics. |
| Approach: | They propose a novel short text topic modeling framework, Topic-Semantic Contrastive Topic Model, which uses augmented data and the data characteristic to learn the relations among samples. |
| Outcome: | The proposed framework outperforms state-of-the-art baselines regardless of the data augmentation availability, producing high-quality topics and topic distributions. |
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. |
CodeArena: A Collective Evaluation Platform for LLM Code Generation (2025.acl-demo)
Copied to clipboard
Mingzhe Du, Anh Tuan Luu, Bin Ji, Xiaobao Wu, Yuhao Qing, Dong Huang, Terry Yue Zhuo, Qian Liu, See-Kiong Ng
| Challenge: | Large Language Models (LLMs) have reshaped code generation, but persistent challenges impede accurate assessment. |
| Approach: | They propose an online evaluation framework tailored for large language models to assess their coding capabilities. |
| Outcome: | a new evaluation framework for large language models (LLMs) provides unbiased, unbiased evaluations and open access to solutions and test cases. |
Modeling Dynamic Topics in Chain-Free Fashion by Evolution-Tracking Contrastive Learning and Unassociated Word Exclusion (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing dynamic topic models lack the ability to reveal the evolution of topics . Existing models suffer from repetitive topic and unassociated topic issues . |
| Approach: | They propose a new evolution-tracking contrastive learning method that builds the similarity relations among dynamic topics and an unassociated word exclusion method to avoid unassociated topics. |
| Outcome: | The proposed model outperforms state-of-the-art models on downstream tasks and is robust to evolution intensities. |
Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failures (2026.acl-long)
Copied to clipboard
Yi Hu, Jiaqi Gu, Ruxin Wang, Zijun Yao, Hao Peng, Xiaobao Wu, Jianhui Chen, Muhan Zhang, Liangming Pan
| Challenge: | Recent research has shown that reinforcement learning can elicit intriguing emergent reasoning behaviors. |
| Approach: | They propose a comprehensive survey of the mechanistic understanding of large reasoning models . they organize findings into three core dimensions: 1) training dynamics, 2) reasoning mechanisms, and 3) unintended behaviors. |
| Outcome: | This paper synthesizes the mechanistic understanding of large reasoning models into three dimensions . authors outline a roadmap for future studies including improved interpretability and methodologies . |
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. |
P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs (2026.findings-acl)
Copied to clipboard
| Challenge: | Defending Large Language Models (LLMs) against backdoors has long been trapped in a "cat-and-mouse" dilemma where defenders passively react to ever-shifting attack strategies. |
| Approach: | They propose a general and effective defense algorithm that implants benign triggers to reshape the model’s decision boundary. |
| Outcome: | The proposed defense algorithm can neutralize malicious backdoors while preserving task performance. |
Discovering Topics in Long-tailed Corpora with Causal Intervention (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing topic models are designed and evaluated on balanced corpora, but the longtailed bias can impair the performance of topic models. |
| Approach: | They propose a causal inference framework to explain topic modeling on long-tailed corpora by applying causal intervention in training. |
| Outcome: | The proposed model can mitigate the bias effect, greatly improve topic quality and discover the hidden semantics on the tail. |
SCOPE: Compress Mathematical Reasoning Steps for Efficient Automated Process Annotation (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing process annotation approaches are computationally expensive. |
| Approach: | They propose a compression-based approach that transforms reasoning steps into code and normalizes them through Abstract Syntax Tree. |
| Outcome: | The proposed method outperforms existing methods on Best-of-N strategy and ProcessBench. |
Towards the TopMost: A Topic Modeling System Toolkit (2024.acl-demos)
Copied to clipboard
| Challenge: | Current topic models adopt totally different datasets, implementations, and evaluations, hindering their research progress and applications. |
| Approach: | They propose a Topic Modeling System Toolkit that covers a broader spectrum of topic modeling scenarios with their complete lifecycles. |
| Outcome: | The proposed toolkit covers a broader spectrum of topic modeling scenarios with their complete lifecycles, including datasets, preprocessing, models, training, and evaluations. |
MUR: Momentum Uncertainty guided Reasoning for Large Language Models (2026.acl-long)
Copied to clipboard
Hang Yan, Fangzhi Xu, Rongman Xu, Yifei Li, Jian Zhang, Haoran Luo, Xiaobao Wu, Anh Tuan Luu, Haiteng Zhao, Qika Lin, Jun Liu
| Challenge: | Existing methods for optimizing reasoning quality are limited by overthinking. |
| Approach: | They propose a method that allocates thinking budgets to critical reasoning steps by tracking and aggregating step-wise uncertainty over time. |
| Outcome: | The proposed method reduces computation by over 45% on average while improving accuracy by 0.33–3.46%. |
AntiLeakBench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge (2025.acl-long)
Copied to clipboard
Xiaobao Wu, Liangming Pan, Yuxi Xie, Ruiwen Zhou, Shuai Zhao, Yubo Ma, Mingzhe Du, Rui Mao, Anh Tuan Luu, William Yang Wang
| Challenge: | Existing studies solve this challenge by updating benchmarks with newly collected data, but they fail to guarantee contamination-free evaluation as the newly collected knowledge may contain pre-existing knowledge. |
| Approach: | They propose an automated anti-leakage benchmarking framework that builds and updates benchmarks without human labor instead of using newly collected data. |
| Outcome: | The proposed framework significantly reduces the cost of benchmark maintenance to accommodate emerging LLMs. |
A Comprehensive Survey on Learning from Rewards for Large Language Models: Reward Models and Learning Strategies (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Recent developments in Large Language Models have shifted from pre-training to post-training and test-time scaling. |
| Approach: | They present a comprehensive overview of learning from rewards from the perspective of reward models and learning strategies across training, inference, and post-inference stages. |
| Outcome: | The proposed paradigm enables the transition from passive learning from static data to active learning from dynamic feedback. |
Three Minds, One Legend: Jailbreak Large Reasoning Model with Adaptive Stacked Ciphers (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing jailbreak methods struggle to balance effectiveness with robustness against adaptive safety mechanisms. |
| Approach: | They propose a novel approach that targets Large Reasoning Models through an adaptive encryption pipeline designed to overwhelm their reasoning capabilities. |
| Outcome: | The proposed approach achieves an attack success rate of 85.6% on OpenAI GPT-o4-mini, outperforming state-of-the-art baselines by a significant margin of 17.2%. |
Zero-Shot Text Classification via Self-Supervised Tuning (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing solutions to zero-shot text classification use pre-trained language models or large-scale annotated data. |
| Approach: | They propose a self-supervised learning paradigm to solve zero-shot text classification tasks by tuning the language models with unlabeled data. |
| Outcome: | The proposed model outperforms the state-of-the-art models on 7 out of 10 tasks and is less sensitive to prompt design. |
Learning Uncertainty from Sequential Internal Dispersion in Large Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Recent approaches to detect hallucinations depend on model internal states to estimate uncertainty, but they focus on last or mean tokens. |
| Approach: | They propose a supervised hallucination detection framework that leverages token-wise, layer-wise features derived from hidden states. |
| Outcome: | The proposed framework outperforms baseline models and avoids large training sets. |
Are LLMs Good Zero-Shot Fallacy Classifiers? (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing fallacy classifiers lack sufficient labeled data for training, limiting their out-of-distribution (OOD) generalization abilities. |
| Approach: | They propose to use Large Language Models (LLMs) for zero-shot fallacy classification. |
| Outcome: | The proposed schemes outperform existing classifiers in OOD inference scenarios and opendomain tasks. |
DemaFormer: Damped Exponential Moving Average Transformer with Energy-Based Modeling for Temporal Language Grounding (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Temporal Language Grounding (TLG) is a task to determine temporal boundaries of video moments that correspond to a language query. |
| Approach: | They propose an energy-based model framework to explicitly learn moment-query distributions. |
| Outcome: | The proposed model outperforms the state-of-the-art models on four public temporal language grounding datasets. |
Enhancing Multimodal Entity Linking with Jaccard Distance-based Conditional Contrastive Learning and Contextual Visual Augmentation (2025.naacl-long)
Copied to clipboard
Cong-Duy T Nguyen, Xiaobao Wu, Thong Thanh Nguyen, Shuai Zhao, Khoi M. Le, Nguyen Viet Anh, Feng Yichao, Anh Tuan Luu
| Challenge: | Existing approaches to multimodal entity linking use contrastive learning to align input sentences and entities, but are limited by their random negative sampling. |
| Approach: | They propose a method to match negative samples with similar attributes using JD-CCL . they also propose 'contextual visual-aid controllable patch transform' experimental results demonstrate the strong effectiveness of their method . |
| Outcome: | The proposed method is able to match negative samples with similar attributes on a multimodal knowledge graph. |
Unlearning Backdoor Attacks for LLMs with Weak-to-Strong Knowledge Distillation (2025.findings-acl)
Copied to clipboard
| Challenge: | Parameter-efficient fine-tuning (PEFT) can bridge the gap between large language models and downstream tasks, but is vulnerable to malicious attacks. |
| Approach: | They propose a weak-to-strong unlearning algorithm based on feature alignment knowledge distillation to defend against backdoor attacks . they first train a small-scale language model through full-parameter fine-tuning to serve as the clean teacher model and then guide the large-scale poisoned student model in unlearning the backdoor. |
| Outcome: | The proposed method can unlearn backdoor features without compromising model performance. |
KDMCSE: Knowledge Distillation Multimodal Sentence Embeddings with Adaptive Angular margin Contrastive Learning (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing work on multimodal sentence embeddings took negative samples without reviewing, resulting in noisy and noisy negative samples. |
| Approach: | They propose a multimodal contrastive learning approach that inherits the knowledge from the teacher model to learn the difference between positive and negative instances. |
| Outcome: | The proposed approach can detect noisy and wrong negative samples before they are calculated in the contrastive objective. |
Zero-to-Strong Generalization: Eliciting Strong Capabilities of Large Language Models Iteratively without Gold Labels (2025.coling-main)
Copied to clipboard
| Challenge: | Pre-trained language models have demonstrated remarkable performance through supervised fine-tuning or in-context learning using gold labels. |
| Approach: | They propose a new paradigm termed zero-to-strong generalization that prompts LLMs to annotate unlabeled data and retain high-quality labels by filtering. |
| Outcome: | The proposed framework outperforms pre-trained language models on extensive classification and reasoning tasks on multiple model sizes. |
Evaluating the Expressive Appropriateness of Speech in Rich Contexts (2026.acl-long)
Copied to clipboard
Tianrui Wang, Ziyang Ma, Yizhou Peng, Haoyu Wang, Zhikang Niu, Zikang Huang, Yihao Wu, Yi-Wen Chao, Yu Jiang, Yuheng Lu, Guanrou Yang, Xuanchen Li, Hexin Liu, Chunyu Qiang, Cheng Gong, Yifan Yang, Tianchi Liu, Junyu Wang, Nana Hou, Meng Ge, Fuming You, Yang Wei, Zhongqian Sun, Hu Haifeng, Xiaobao Wang, Eng Siong Chng, Xie Chen, Longbiao Wang, Jianwu Dang
| Challenge: | Existing methods for evaluating expressive speech focus on word accuracy, naturalness, signal quality, or emotional intensity at the utterance level. |
| Approach: | They propose a framework for Evaluating Expressive Appropriateness in speech that assesses whether a speech sample aligns with the underlying communicative intent implied by its discourse-level narrative context. |
| Outcome: | The proposed framework outperforms existing speech evaluation and analysis systems on a human-annotated test set. |
Short Text Topic Modeling with Topic Distribution Quantization and Negative Sampling Decoder (2020.emnlp-main)
Copied to clipboard
| Challenge: | Topic models for short texts suffer from data sparsity because of limited word co-occurrences. |
| Approach: | They propose a neural topic model with a new topic distribution quantization approach that generates peakier distributions that are more appropriate for modeling short texts. |
| Outcome: | The proposed model outperforms both strong traditional and neural baselines under extreme data sparsity scenes, producing high-quality topics. |
RuleArena: A Benchmark for Rule-Guided Reasoning with LLMs in Real-World Scenarios (2025.acl-long)
Copied to clipboard
| Challenge: | RuleArena assesses the ability of large language models (LLMs) to follow complex, real-world rules in reasoning. |
| Approach: | They propose a benchmark to evaluate the ability of large language models (LLMs) to follow complex, real-world rules in reasoning. |
| Outcome: | The proposed benchmark covers airline baggage fees, NBA transactions, and tax regulations. |
AKEW: Assessing Knowledge Editing in the Wild (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent Large Language Models (LLMs) have revolutionized the NLP field but their knowledge could become incorrect or outdated over time. |
| Approach: | They propose a new practical benchmark for knowledge editing that covers three editing settings of knowledge updates: structured facts, unstructured texts as facts, and extracted triplets. |
| Outcome: | The proposed method covers structured facts, unstructured texts as facts, and extracted triplets. |
MRAG: A Modular Retrieval Framework for Time-Sensitive Question Answering (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for answering time-sensitive questions lack temporal reasoning . existing methods struggle with these time-intensive questions, authors say . |
| Approach: | They propose a temporal-based question-answering framework that integrates temporal perturbations and gold evidence labels into a question processing framework. |
| Outcome: | The proposed framework outperforms baseline retrieval methods in retrieval performance. |