Papers by Weizhi Zhang
TestNUC: Enhancing Test-Time Computing Approaches and Scaling through Neighboring Unlabeled Data Consistency (2025.acl-long)
Copied to clipboard
Henry Peng Zou, Zhengyao Gu, Yue Zhou, Yankai Chen, Weizhi Zhang, Liancheng Fang, Yibo Wang, Yangning Li, Kay Liu, Philip S. Yu
| Challenge: | Test-time computing approaches that leverage additional computational resources during inference have been proven effective in enhancing large language model performance. |
| Approach: | They propose a linearly scaling approach that leverages local consistency of neighboring unlabeled data to improve test-time predictions. |
| Outcome: | The proposed approach outperforms baseline methods such as prompting and self-consistency across eight datasets and performs robustly across embedding models. |
A User-Centric Multi-Intent Benchmark for Evaluating Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing benchmarks focus on specific predefined model abilities, such as world knowledge, reasoning, etc., making it difficult for users to determine which LLM best suits their particular needs. |
| Approach: | They propose to evaluate large language models from a user-centric perspective and use real-world use cases to identify their effectiveness under distinct intents. |
| Outcome: | The proposed benchmarks achieve a correlation between human preference and the user-reported scenarios and human intents. |
TABGEN-ICL: Residual-Aware In-Context Example Selection for Tabular Data Generation (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to tabular data generation require fine-tuning, which is computationally expensive. |
| Approach: | They propose a new in-context learning framework to prompt a fixed LLM with in-constitut examples to enhance the in-text learning ability of LLMs for tabular data generation. |
| Outcome: | The proposed framework outperforms random selection strategies on five real-world tabular datasets and reduces error rate by 42.2% on fidelity metric. |
LLM-Based Human-Agent Collaboration and Interaction Systems: A Survey (2026.findings-acl)
Copied to clipboard
Henry Peng Zou, Wei-Chieh Huang, Yaozu Wu, Jizhou Guo, Yankai Chen, Chunyu Miao, Hoang H Nguyen, Yue Zhou, Weizhi Zhang, Liancheng Fang, Hanrong Zhang, Fangxin Wang, Pengfei Zhang, Langzhou He, Yangning Li, Dongyuan Li, Renhe Jiang, Philip S. Yu
| Challenge: | Recent advances in large language models (LLMs) have sparked growing interest in building fully autonomous agents. |
| Approach: | They propose to integrate human-provided information, feedback, or control into the agent system to enhance system performance, reliability, and safety. |
| Outcome: | The proposed systems improve system performance, reliability, and safety by integrating human-provided information, feedback, or control into the agent system. |
LLMInit: A Free Lunch from Large Language Models for Selective Initialization of Recommendation (2025.emnlp-industry)
Copied to clipboard
Weizhi Zhang, Liangwei Yang, Wooseong Yang, Henry Peng Zou, Yuqing Liu, Ke Xu, Sourav Medya, Philip S. Yu
| Challenge: | Existing algorithms for collaborative filtering are limited by their computational demands and latency. |
| Approach: | They propose a framework to integrate pre-trained LLM embeddings into CF models through selective initialization strategies. |
| Outcome: | The proposed framework improves recommendation performance while maintaining low computational costs. |
A Survey of RAG-Reasoning Systems in Large Language Models (2025.findings-emnlp)
Copied to clipboard
Yangning Li, Weizhi Zhang, Yuyao Yang, Wei-Chieh Huang, Yaozu Wu, Junyu Luo, Yuanchen Bei, Henry Peng Zou, Xiao Luo, Yusheng Zhao, Chunkit Chan, Yankai Chen, Zhongfen Deng, Yinghui Li, Hai-Tao Zheng, Dongyuan Li, Renhe Jiang, Ming Zhang, Yangqiu Song, Philip S. Yu
| Challenge: | a survey of RAG-based reasoning-based approaches shows that it is not effective for multi-step inferences. |
| Approach: | They map how advanced reasoning optimizes each stage of RAG . they show how retrieved knowledge supply missing premises and expand context for complex inference . |
| Outcome: | The proposed frameworks achieve state-of-the-art across knowledge-intensive benchmarks. |
How Far Can LLMs Improve from Experience? Measuring Test-Time Learning Ability in LLMs with Human Comparison (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing benchmarks primarily assess static knowledge, while intelligence also entails the ability to rapidly learn from experience. |
| Approach: | They propose to use semantic games to evaluate test-time learning . they recruit eight human participants to complete the same task . |
| Outcome: | The proposed framework compares model performance under limited and cumulative experience settings and contains four forms of experience representation. |
Many-Shot Scaling of In-Context Learning with Self-Generated Demonstrations (2026.findings-acl)
Copied to clipboard
| Challenge: | In-context learning methods that use self-generated annotations do not scale to many-shot scenarios. |
| Approach: | They propose a framework analogous to semi-supervised learning that uses self-generated annotations instead of ground truth labels. |
| Outcome: | The proposed framework outperforms ground truth ICL under zero-shot, few-shot and many-shot settings. |
GAM: Hierarchical Graph-based Agentic Memory for LLM Agents (2026.acl-long)
Copied to clipboard
Zhaofen Wu, Hanrong Zhang, Fulin Lin, Wujiang Xu, Xinran Xu, Yankai Chen, Henry Peng Zou, Shaowen Chen, Weizhi Zhang, Xue Liu, Philip S. Yu, Hongwei Wang
| Challenge: | Current unified stream-based memory systems facilitate context updates but remain vulnerable to interference from transient noise. |
| Approach: | They propose a hierarchical Graph-based Agentic Memory framework that explicitly decouples memory encoding from consolidation to resolve conflict between rapid context perception and stable knowledge retention. |
| Outcome: | The proposed framework outperforms state-of-the-art benchmarks on LoCoMo and LongDialQA. |
DeCrisisMB: Debiased Semi-Supervised Learning for Crisis Tweet Classification via Memory Bank (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies utilize social media platforms such as Twitter to build models for crisis event analysis, but semi-supervised approaches require annotating vast amounts of data and are impractical due to limited response time. |
| Approach: | They propose a method that stores and performs equal sampling for generated pseudo-labels from each class at each training iteration. |
| Outcome: | The proposed method performs better than existing methods in both in-distribution and out-of-difference settings. |
WebAgent-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning (2025.emnlp-main)
Copied to clipboard
Zhepei Wei, Wenlin Yao, Yao Liu, Weizhi Zhang, Qin Lu, Liang Qiu, Changlong Yu, Puyang Xu, Chao Zhang, Bing Yin, Hyokun Yun, Lihong Li
| Challenge: | Existing work on reinforcement learning has focused on single-turn tasks such as solving math problems. |
| Approach: | They propose a framework that learns directly from online interactions by asynchronously generating diverse trajectories, guided by binary rewards depending on task success. |
| Outcome: | Experiments on the WebArena-Lite benchmark show that the framework outperforms state-of-the-art methods and strong proprietary models. |
SciCustom: A Framework for Custom Evaluation of Scientific Capabilities in Large Language Models (2026.acl-long)
Copied to clipboard
Yiyang Gu, Junwei Yang, Junyu Luo, Ye Yuan, Bin Feng, Yingce Xia, Shufang Xie, Kaili Liu, Bohan Wu, Qi Shi, Haoran Li, Beier Xiao, Zhiping Xiao, Xiao Luo, Weizhi Zhang, Philip S. Yu, Zequn Liu, Ming Zhang
| Challenge: | Existing evaluations of large language models fail to reflect fine-grained capabilities . existing benchmarks are manually curated or domain-generic, limiting scalability and alignment with real use cases. |
| Approach: | They propose a framework that allows custom construction of benchmarks from large-scale scientific data to evaluate application-specific scientific capabilities in LLMs. |
| Outcome: | The proposed framework reveals fine-grained differences in scientific capabilities that standard benchmarks overlook . it allows custom construction of benchmarks from large-scale scientific data to evaluate application-specific capabilities in LLMs. |
Multifaceted Evaluation of Audio-Visual Capability for MLLMs: Effectiveness, Efficiency, Generalizability and Robustness (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Multi-modal large language models have been used for processing and understanding information from diverse modalities. |
| Approach: | They propose to evaluate the audio-visual capabilities of multi-modal large language models . they focus on effectiveness, efficiency, generalizability, and robustness . |
| Outcome: | The proposed models exhibit strong zero-shot and few-shot generalization abilities . their success relies heavily on the vision modality, which impairs performance when visual input is corrupted or missing. |
ImplicitAVE: An Open-Source Dataset and Multimodal LLMs Benchmark for Implicit Attribute Value Extraction (2024.findings-acl)
Copied to clipboard
Henry Zou, Vinay Samuel, Yue Zhou, Weizhi Zhang, Liancheng Fang, Zihe Song, Philip Yu, Cornelia Caragea
| Challenge: | Existing datasets for attribute value extraction focus on explicit attribute values while neglecting the implicit ones. |
| Approach: | They present a multimodal dataset for implicit attribute value extraction that includes AVE and multimodality. |
| Outcome: | The proposed dataset includes 68k training and 1.6k testing data across five domains. |
PersonaAgent: Bridging Memory and Action for Personalized LLM Agents (2026.findings-acl)
Copied to clipboard
Weizhi Zhang, Xinyang Zhang, Chenwei Zhang, Liangwei Yang, Jingbo Shang, Zhepei Wei, Henry Peng Zou, Zijie Huang, Zhengyang Wang, Yifan Gao, Xiaoman Pan, Lian Xiong, Jingguo Liu, Philip S. Yu, Xian Li
| Challenge: | Existing Large Language Model (LLM) enabled agents lack flexibility to respond to users’ varying needs and preferences. |
| Approach: | They propose a test-time user-preference alignment strategy that optimizes the persona prompt, ensuring real-time preference alignment through textual loss feedback between simulated and ground-truth responses. |
| Outcome: | The proposed framework outperforms baseline methods in real-time and in real applications. |
Non-Parametric Domain Adaptation for End-to-End Speech Translation (2022.emnlp-main)
Copied to clipboard
| Challenge: | End-to-end speech translation (E2E-ST) systems have received increasing attention due to its less error propagation, lower latency and fewer parameters. |
| Approach: | They propose a non-parametric method that leverages in-domain text translation corpus to achieve domain adaptation for E2E-ST systems. |
| Outcome: | The proposed method outperforms the existing in-domain fine-tuning strategies on the Europarl-ST benchmark. |
Efficient Dynamic Clustering-Based Document Compression for Retrieval-Augmented-Generation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | RAG implementations face challenges in addressing retrieved noise and redundant content . current RAG methods lack the ability to exploit fine-grained inter-document relationships . |
| Approach: | They propose a retrieval-augmented generation framework that exploits latent inter-document relationships while removing irrelevant information and redundant content. |
| Outcome: | The proposed framework achieves consistent performance improvements on knowledge-QA and hallucination-Detection datasets. |
Rethinking Zero-shot Neural Machine Translation: From a Perspective of Latent Variables (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to achieve zero-shot translation suffer from spurious correlations between output language and language invariant semantics. |
| Approach: | They propose a method that denoizes the autoencoder objective based on pivot language into traditional training objective to improve translation accuracy on zero-shot directions. |
| Outcome: | The proposed method eliminates spurious correlations and outperforms state-of-the-art methods on two benchmark machine translation datasets. |
Deep Research with Open-Domain Evaluation and Multi-Stage Guardrails for Safety (2026.acl-long)
Copied to clipboard
Wei-Chieh Huang, Henry Peng Zou, Yaozu Wu, Dongyuan Li, Yankai Chen, Weizhi Zhang, Yangning Li, Angelo Zangari, Jizhou Guo, Chunyu Miao, Liancheng Fang, Langzhou He, Yinghui Li, Renhe Jiang, Philip S. Yu
| Challenge: | Existing deep research frameworks lack adequate evaluation procedures and stage-specific protections. |
| Approach: | They propose a framework with open-domain evaluation and a stage-wise safety benchmark to address this oversight. |
| Outcome: | The proposed framework improves defense success rates by 16.53% while reducing over-refusal rates to approximately 6%. |