Papers by Yong Hu
Towards Reward Fairness in RLHF: From a Resource Allocation Perspective (2025.acl-long)
Copied to clipboard
| Challenge: | if rewards are imperfect, they can adversely affect the alignment of large language models (LLMs). |
| Approach: | They propose a bias-agnostic method to address the issue of reward unfairness from a resource allocation perspective without specifically designing for each type of bias . they apply methods Fairness Regularization and Fairness Coefficient to achieve fairness in rewards. |
| Outcome: | The proposed method achieves fairness in rewards while minimizing biases . it can be applied to verification and reinforcement learning scenarios . |
Risk Minimization for Zero-shot Sequence Labeling (2021.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to zero-shot sequence labeling are expensive and hard to obtain for lowresource languages/domains. |
| Approach: | They propose a framework for zero-shot sequence labeling with minimum risk training and a decomposable risk function that models the relations between predicted labels from the source models and the true labels. |
| Outcome: | The proposed framework outperforms state-of-the-art systems on 21 datasets. |
Encouraging Good Processes Without the Need for Good Answers: Reinforcement Learning for LLM Agent Planning (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Currently, the dominant end-to-end reinforcement learning paradigm for agents in Large Language Models (LLMs) employs multi-objective optimization that jointly trains both planning and answer summarization capabilities. |
| Approach: | They propose a framework that decouples the training process to enable a focused, single-objective optimization of the planning module. |
| Outcome: | The proposed framework achieves an 8%–12% improvement in planning performance compared to end-to-end baselines. |
CityEQA: A Hierarchical LLM Agent on Embodied Question Answering Benchmark in City Space (2025.emnlp-main)
Copied to clipboard
Yong Zhao, Kai Xu, Zhengqiu Zhu, Yue Hu, Zhiheng Zheng, Yingfeng Chen, Yatai Ji, Chen Gao, Yong Li, Jincai Huang
| Challenge: | Embodied Question Answering (EQA) tasks are primarily focused on indoor environments, leaving the complexities of urban settings unexplored. |
| Approach: | They propose a task where an embodied agent answers open-vocabulary questions in dynamic city spaces. |
| Outcome: | The proposed agent achieves 60.7% of human-level answering accuracy compared to baselines . the proposed agent outperforms existing agents in open-ended city spaces . |
Mitigating the Alignment Tax of RLHF (2024.emnlp-main)
Copied to clipboard
Yong Lin, Hangyu Lin, Wei Xiong, Shizhe Diao, Jianmeng Liu, Jipeng Zhang, Rui Pan, Haoxiang Wang, Wenbin Hu, Hanning Zhang, Hanze Dong, Renjie Pi, Han Zhao, Nan Jiang, Heng Ji, Yuan Yao, Tong Zhang
| Challenge: | Large Language Models (LLMs) acquire a wide range of abilities during pre-training, but aligning LLMs under Reinforcement Learning with Human Feedback (RLHF) can lead to forgetting pretrained abilities, which is also known as the alignment tax. |
| Approach: | They propose to use a model averaging technique to find the most powerful alignment-forging Pareto front among RLHF algorithms. |
| Outcome: | The proposed method achieves the strongest alignment-forging Pareto front among competing methods. |
Explore More Guidance: A Task-aware Instruction Network for Sign Language Translation Enhanced with Data Augmentation (2022.findings-naacl)
Copied to clipboard
| Challenge: | Existing studies focus on the recognition step, while paying less attention to sign language translation. |
| Approach: | They propose a task-aware instruction network, namely TIN-SLT, for sign language translation, by introducing the isntruction module and the learning-based feature fuse strategy into a Transformer network. |
| Outcome: | The proposed system outperforms existing solutions on two benchmark datasets, PHOENIX-2014-T and ASLG-PC12, and outperformed previous best solutions by 1.65 and 1.42 in terms of BLEU-4. |
Adversarial Preference Optimization: Enhancing Your Alignment via RM-LLM Game (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for training large language models require additional annotations to adjust to shifted distributions. |
| Approach: | They propose an algorithm that allows LLMs and reward models to update alternatively via a min-max game to improve their alignment. |
| Outcome: | The proposed framework improves existing alignment baselines in terms of LLM helpfulness and harmlessness. |
Domain-Specific NER via Retrieving Correlated Samples (2022.coling-1)
Copied to clipboard
| Challenge: | Successful Named Entity Recognition models fail on texts from some special domains, for example, Chinese addresses and e-commerce titles. |
| Approach: | They propose to enhance NER models with correlated samples to help the text understanding . they draw correlated texts by the sparse BM25 retriever from large-scale in-domain unlabeled data . |
| Outcome: | Empirical results show that NER models can be enhanced with correlated samples . the proposed model can be used to reason out the correct answer on hard cases . |
Visually-Guided Policy Optimization for Multimodal Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing RLVRs lack visual faithfulness due to text-dominated reasoning . a novel framework to reinforce visual focus during policy optimization is proposed . |
| Approach: | They propose a framework to reinforce visual focus during policy optimization using visual attention compensation mechanism. |
| Outcome: | The proposed framework exhibits better visual activation and superior performance in multimodal reasoning and visual-dependent tasks. |
KBM: Delineating Knowledge Boundary for Adaptive Retrieval in Large Language Models (2025.findings-emnlp)
Copied to clipboard
Zhen Zhang, Xinyu Wang, Yong Jiang, Zile Qiao, Zhuo Chen, Guangyu Li, Feiteng Mu, Mengting Hu, Pengjun Xie, Fei Huang
| Challenge: | Retrieval-augmented generation (RAG) is employed to tackle these challenges . a Knowledge Boundary Model (KBM) is used to express the known/unknown of a given question . |
| Approach: | They propose a Knowledge Boundary Model to express the known/unknown of a given question . they find that not all questions need to trigger RAG to improve performance . |
| Outcome: | The proposed model reduces time and computational costs by retrieving parts of unknown knowledge . the proposed model can express the known/unknown of a given question and determine whether a RAG needs to be triggered . |
Entity-to-Text based Data Augmentation for various Named Entity Recognition Tasks (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing augmentation techniques manipulate words in the original text that break the semantic coherence of the text, or exploit generative models that ignore preserving entities in the text. |
| Approach: | They propose a novel Entity-to-Text based data augmentation technique called EnTDA to add, delete, replace or swap entities in the original text. |
| Outcome: | The proposed technique generates semantically coherent and entity preserving texts on thirteen NER tasks and two settings. |
ChatSOP: An SOP-Guided MCTS Planning Framework for Controllable LLM Dialogue Agents (2025.acl-long)
Copied to clipboard
Zhigen Li, Jianxiang Peng, Yanmeng Wang, Yong Cao, Tianhao Shen, Minghui Zhang, Linxi Su, Shang Wu, Yihang Wu, YuQian Wang, Ye Wang, Wei Hu, Jianfeng Li, Shaojun Wang, Jing Xiao, Deyi Xiong
| Challenge: | Existing models that use Large Language Models (LLMs) show superior performance in various tasks, but lack of controllability leads to unfocused conversations or task failure. |
| Approach: | They propose a standard operating procedure (SOP) framework to regulate dialogue flow by integrating Chain of Thought reasoning and supervised fine-tuning for SOP prediction. |
| Outcome: | The proposed method achieves a 27.95% improvement in action accuracy compared to baseline models based on GPT-3.5 and also shows notable gains for open-source models. |
Attribution-Based Analysis and Optimization of Modular Agentic Workflows (2026.findings-acl)
Copied to clipboard
Yingxuan Yang, Bo Huang, Siyuan Qi, Chao Feng, Haoyi Hu, Yuxuan Zhu, Jinbo Hu, Haoran Zhao, Ziyi He, Xiao Liu, ZongYu Wang, Muning Wen, Lin Qiu, Xuezhi Cao, Xunliang Cai, Yong Yu, Weinan Zhang
| Challenge: | Large Language Models (LLMs) have driven the rise of agentic workflows . yet, how can we attribute performance gains to individual upgrades and their interactions? |
| Approach: | They propose a game-theoretic framework that models component upgrades as players and evaluates component coalitions to compute Shapley values. |
| Outcome: | The proposed framework provides interaction-aware attribution and recommendation for model allocation under a fixed workflow structure. |
No More Stale Feedback: Co-Evolving Critics for Open-World Agent Learning (2026.acl-long)
Copied to clipboard
Zhicong Li, Lingjie Jiang, Yulan Hu, Xingchen Zeng, Yixia Li, Xiangwen Zhang, Guanhua Chen, Zheng Pan, Xin Li, Yong Liu
| Challenge: | Current methods for training Large Language Model agents rely on static or offline critic models, which fail to adapt as the policy evolves. |
| Approach: | They propose a framework that integrates a critique and a policy to optimize the policy and critic through a synchronized co-evolutionary loop. |
| Outcome: | The proposed framework yields more stable training and higher long-horizon task success across open-world environments. |
On Diversified Preferences of Large Language Model Alignment (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models (LLMs) can be fine tuned with human feedback, but human preferences can be diversified due to annotators’ different tastes, which hinders the effectiveness of LLM alignment methods. |
| Approach: | They propose a calibration error metric to evaluate large language models (LLMs) and a multi-objective reward learning method to enhance the calibration performance of RMs on shared preferences. |
| Outcome: | The proposed model can be adopted as a key calibration error and MORE can achieve superior alignment performance. |
Thinking with Map: Reinforced Parallel Map-Augmented Agent for Geolocalization (2026.findings-acl)
Copied to clipboard
Yuxiang Ji, Yong Wang, Ziyu Ma, Yiming Hu, Hailang Huang, Xuecai Hu, Guanhua Chen, Liaoni Wu, Xiangxiang Chu
| Challenge: | Existing large vision-language model (LVLM) approaches overlook a common strategy used by humans — using maps. |
| Approach: | They propose a method to equip a vision-language model with the ability to think with maps and optimize it using agentic reinforcement learning and parallel test-time scaling. |
| Outcome: | The proposed method outperforms open- and closed-source models on most metrics. |
GOLD: Generalized Knowledge Distillation via Out-of-Distribution-Guided Language Data Generation (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods to generate data using LLMs are limited by sampling from the center of original content distribution. |
| Approach: | They propose a task-agnostic data generation and knowledge distillation framework for LLMs that employs an iterative out-of-distribution-guided feedback mechanism to generate data. |
| Outcome: | The proposed framework outperforms prior arts and the LLM on 10 different classification tasks and noisey generated data. |
CityCube: Benchmarking Cross-view Spatial Reasoning on Vision-Language Models in Urban Environments (2026.acl-long)
Copied to clipboard
Haotian Xu, Yue Hu, Zhengqiu Zhu, Chen Gao, Ziyou Wang, Junreng Rao, Wenhao Lu, Weishi Li, Quanjun Yin, Yong Li
| Challenge: | Existing benchmarks focus on indoor or street settings, overlooking challenges of open-ended urban spaces. |
| Approach: | They propose a benchmark to probe cross-view spatial reasoning capabilities of current VLMs in urban settings. |
| Outcome: | The citycube benchmark examines the performance of current vision-language models in urban environments. |
C-LLM: Learn to Check Chinese Spelling Errors Character by Character (2024.emnlp-main)
Copied to clipboard
| Challenge: | Chinese Spell Checking (CSC) aims to detect and correct spelling errors in sentences. |
| Approach: | They propose a Chinese Spell Checking method that learns to check errors Character by Character. |
| Outcome: | The proposed method achieves a 2.1% enhancement in general scenarios and a significant improvement in vertical domain scenarios compared to existing methods. |
Integrating Semantic Scenario and Word Relations for Abstractive Sentence Summarization (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing graph-based methods only consider word relations or structure information, which neglect the correlation between them. |
| Approach: | They propose a Dual Graph network for Abstractive Sentence Summarization that captures word relations and structure information from sentences. |
| Outcome: | The proposed model outperforms state-of-the-art methods on two popular benchmark datasets. |
Multi-View Cross-Lingual Structured Prediction with Minimum Supervision (2021.acl-long)
Copied to clipboard
| Challenge: | Existing work on cross-lingual transfer learning focuses on transferring knowledge from high-resource languages to low-resourced ones. |
| Approach: | They propose a multi-view framework that integrates multiple source models into an aggregated source view and transfers it to a target view based on a task-specific model. |
| Outcome: | The proposed framework improves on three structured prediction tasks on 16 datasets. |
CoEvolve: Training LLM Agents via Agent-Data Mutual Evolution (2026.acl-long)
Copied to clipboard
| Challenge: | Extensive experiments on AppWorld and BFCL demonstrate consistent and significant improvements over strong base models, yielding absolute gains of 19.43%, 15.58%, and 18.14%, respectively. |
| Approach: | They propose a framework that extracts feedback signals such as forgetting and uncertainty from rollout trajectories and utilizes them to guide LLM-based task synthesis. |
| Outcome: | Extensive experiments on AppWorld and BFCL show that the proposed framework improves over strong base models. |
CANDY: Benchmarking LLMs’ Limitations and Assistive Potential in Chinese Misinformation Fact-Checking (2025.findings-emnlp)
Copied to clipboard
| Challenge: | CANDY is a benchmark to evaluate the capabilities and limitations of large language models (LLMs) for fact-checking misinformation. |
| Approach: | a team of researchers develop a benchmark to evaluate the capabilities and limitations of large language models in fact-checking misinformation in Chinese. |
| Outcome: | CANDY is a benchmark to evaluate the capabilities and limitations of large language models in fact-checking misinformation in China. |
Adaptive Threshold Selective Self-Attention for Chinese NER (2022.coling-1)
Copied to clipboard
| Challenge: | Named entity recognition (NER) is a computationally difficult task in Chinese since there is no natural delimiter between words in sentences. |
| Approach: | They propose a data-driven Adaptive Threshold Selective Self-Attention mechanism to select the most relevant characters to enhance Transformer architecture for Chinese named entity recognition. |
| Outcome: | Experiments on four benchmark Chinese NER datasets show the proposed mechanism improves performance. |
HVGuard: Utilizing Multimodal Large Language Models for Hateful Video Detection (2025.emnlp-main)
Copied to clipboard
Yiheng Jing, Mingming Zhang, Yong Zhuang, Jiacheng Guo, Juan Wang, Xiaoyang Xu, Wenzhe Yi, Keyan Guo, Hongxin Hu
| Challenge: | Existing methods for hateful video detection rely on unimodal analysis or feature fusion . Existing tools struggle to capture cross-modal interactions and reason through implicit hate in sarcasm and metaphor . |
| Approach: | They propose a reasoning-based hateful video detection framework with multimodal large language models . they integrate Chain-of-Thought reasoning to enhance multimodal interaction modeling . |
| Outcome: | The proposed framework outperforms existing tools on two public datasets covering English and Chinese. |
SPPD: Self-training with Process Preference Learning Using Dynamic Value Margin (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to improve numerical and logical reasoning of Large Language Models are limited . existing approaches rely on prompt engineering and pretrained knowledge to ensure correctness . |
| Approach: | They propose to train LLMs with process-based reasoning using a dynamic value margin . they use the Bellman optimality equation to derive a value margin for step-level preference optimization . |
| Outcome: | The proposed method is equivalent to on-policy policy gradient methods under constrained reward functions. |
CSCD-NS: a Chinese Spelling Check Dataset for Native Speakers (2024.acl-long)
Copied to clipboard
| Challenge: | Existing datasets aimed at Chinese learners and native speakers are limited in size and quality. |
| Approach: | They propose a method that simulates the input process through an input method and generates large-scale pseudo data that closely resembles the actual error distribution. |
| Outcome: | The proposed method outperforms existing methods and outperformed existing models. |
An Investigation of Potential Function Designs for Neural CRF (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to sequence labeling are based on the neural linear-chain CRF model. |
| Approach: | They propose a series of increasingly expressive potential functions for neural CRF models that integrate emission and transition functions and explicitly take contextual words as input. |
| Outcome: | The proposed model consistently achieves the best performance on the decomposed quadrilinear potential function based on the representations of two neighboring labels and two neighbored words. |
Beyond Itinerary Planning—A Real-World Benchmark for Multi-Turn and Tool-Using Travel Tasks (2026.acl-long)
Copied to clipboard
| Challenge: | Existing studies on LLM performance on travel planning have shown that existing settings are limited due to limited domain coverage, insufficient modeling of users’ implicit preferences in multi-turn conversations, and a lack of evaluation of agents’ capability boundaries. |
| Approach: | They propose a benchmark to evaluate LLMs' planning and tool-use abilities in real-world settings by collecting user queries, user preferences, and tools from real scenarios. |
| Outcome: | The proposed benchmark evaluates agents' capabilities in real-world settings and shows that even advanced models exhibit imbalanced performance across different capabilities. |
Partial Order-centered Hyperbolic Representation Learning for Few-shot Relation Extraction (2025.coling-main)
Copied to clipboard
| Challenge: | Existing methods for few-shot relation extraction are limited to labeled instances and rely on data labeling. |
| Approach: | They propose a partial order-centered hyperbolic representation learning framework which imposes constraints on relations on instances by modeling partial order in hyperbolical space. |
| Outcome: | The proposed framework outperforms baseline methods on three benchmark datasets on 1-shot settings lacking relation descriptions. |