Email: chuanx [at] nagoya-u.jp

Phone: +81-6-6105-6502 

Address:

Graduate School of Information Science and Technology

Osaka University

1-5, Yamadaoka, Suita, Osaka 565-0871

Japan

[DBLP Entry] [Google Scholar]

News

• Truth-Shapley: We investigated data overvaluation attack in federated learning and proposed a data valuation metric as a solution to truthful data valuation. 

• SAS: We are excited to announce the launch of SAS (Smart Agent Survey), a software designed for automated surveys utilizing LLM agents. Users can upload their questionnaires in various formats (DOCX, PDF, etc.). The software employs LLM agents to simulate respondent answers, making it a valuable tool for marketing strategies and policy-making. 

• Ultraverse: We developed a framework for efficient what-if analysis in database-intensive web applications. Our paper will be presented at SIGMOD 2025.

• SIMformer: We discovered that a single-layer vanilla transformer can learn representations to capture trajectory similarity, and the performance is excellent! Our paper is to appear at PVLDB. [source code]

• PrivBench: We developed a privacy-preserving database synthesis method for benchmarking publishing. Our paper is to appear at PVLDB. [source code]

• Our case study of plea bargaining has been accepted to WITS 2024.

• LFP: We introduced the task of legal fact prediction and constructed a dataset for civil loan cases. Experiments were conducted using GPT-4o agents. Benchmark is here.

• Our tutorial On the Use of Large Language Models for Table Tasks was presented at CIKM 2024.

• LLMob: Our paper, introducing an LLM agent framework for the generation of personal activity trajectories, is to appear at NeurIPS 2024. [source code] 

• Jellyfish: As a joint work with NEC, we released the 7B, 8B, and 13B models for data preprocessing, by fine-tuning Mistral-7B-Instruct-v0.2, Llama-3-8B-Instruct, and OpenOrca-Platypus2-13B, respectively. We also released our instruction data used to build these models. Our paper appeared at EMNLP 2024. [paper] [slides] 

• Shall We Team Up: Following the SABM framework, we discovered the spontaneous cooperation of LLM agents in competing environments. Our paper appeared at EMNLP 2024 Findings. [slides] [source code]