Zefang Liu, Yinzhu Quan
arXiv preprint arXiv:2506.08136, 2025
EconWebArena is a benchmark of 360 human-curated tasks from real websites that evaluates autonomous agents’ ability to perform complex, multimodal economic reasoning and web navigation using authoritative sources.
Zefang Liu
2025 13th International Symposium on Digital Forensics and Security (ISDFS), 2025
AutoBnB uses LLM-based agents with the Backdoors & Breaches framework to simulate and evaluate multi-agent collaboration strategies for cybersecurity incident response.
Zefang Liu
AAAI 2025 Workshop on Multi-Agent AI in the Real World (MARW), 2024
This paper explores LLM-based multi-agent collaboration in incident response, analyzing team structures using the Backdoors & Breaches card game.
Zefang Liu, Yinzhu Quan
NAACL 2025 Workshop on Knowledge-Augmented Methods for NLP (KnowledgeNLP), 2024
We present TESRBench, a benchmark for retrieving temporal event sequences from descriptions, and TPP-Embedding, a model aligning sequences and descriptions in a shared embedding space, achieving state-of-the-art performance on TESRBench.
Zefang Liu, Yinzhu Quan
ICLR 2025 Workshop on Foundation Models in the Wild (FM-Wild), 2024
This paper introduces TPP-LLM, a framework that integrates large language models with temporal embeddings to improve event sequence modeling and prediction, demonstrating strong performance across multiple real-world datasets.
Zefang Liu, Weston M. Stacey
AAAI 2025 Workshop on AI to Accelerate Science and Engineering (AI2ASE), 2024
A neural ODE-based burning plasma dynamics model, NeuralPlasmaODE, simulates energy transfer in ITER plasmas for both inductive and non-inductive scenarios.
Yinzhu Quan, Zefang Liu
AAAI 2025 Workshop on Advancing LLM-Based Multi-Agent Collaboration (WMAC), 2024
InvAgent leverages large language models for adaptive and explainable multi-agent inventory management in supply chains, significantly improving efficiency and resilience.
Yinzhu Quan, Zefang Liu
2024 Conference on Empirical Methods in Natural Language Processing (EMNLP) Findings, 2024
We introduce EconLogicQA, a benchmark designed to evaluate large language models’ ability to understand and sequence complex economic events, demonstrating its effectiveness through evaluations of various models.
Zefang Liu, Jiahua Luo
2024 Conference on Language Modeling (COLM), 2024
AdaMoLE introduces a dynamic approach to fine-tuning LLMs by using an adaptive mixture of LoRA experts, outperforming traditional top-k routing methods in various tasks.
Zefang Liu, Weston M. Stacey
ICLR 2024 Workshop on AI4DifferentialEquations in Science (AI4DiffEqtnsInSci), 2024
A novel Neural ODE-based model is introduced for simulating tokamak plasma dynamics, offering precise energy transfer analysis crucial for advancing controlled thermonuclear fusion.
Zefang Liu
2024 12th International Symposium on Digital Forensics and Security (ISDFS), 2024
This paper examines how pre-trained language models enhance cybersecurity across various tasks and calls for continued innovation in this field.
[Paper]
Zefang Liu, Jialei Shi, John F. Buford
AAAI 2024 Workshop on Artificial Intelligence for Cyber Security (AICS), 2024
CyberBench introduces a benchmark for evaluating large language models in cybersecurity, alongside CyberInstruct, a fine-tuned LLM that performs competitively in this domain.
Zefang Liu
arXiv preprint arXiv:2312.15838, 2023
SecQA introduces a specialized dataset for evaluating the performance of large language models in computer security through multiple-choice questions.
Zefang Liu, John F. Buford
2023 Conference on Applied Machine Learning in Information Security (CAMLIS), 2023
This paper leverages DistilBERT for anomaly detection in Unix shell sessions using both unsupervised and supervised methods, demonstrating effective detection of anomalous behavior with minimal data labeling.
Zefang Liu, Weston M. Stacey
64th Annual Meeting of the APS Division of Plasma Physics, 2022
This paper develops a multi-region, multi-timescale transport model to simulate burning plasma dynamics in tokamaks, addressing energy transport and radiation effects to prevent thermal runaway instability in ITER scenarios.
Zefang Liu, Shuran Wen, Yinzhu Quan
arXiv preprint arXiv:2106.06900, 2021
This paper introduces a Deep Reinforcement learning-based Group Recommender System (DRGR) using actor-critic networks and the deep deterministic policy gradient algorithm.
Zefang Liu
arXiv preprint arXiv:2012.06690, 2020
This paper predicts Yelp restaurant ratings using both traditional machine learning models and transformer-based models.