|
Kobe International Workshop on Trustworthy Foundation Models (KTFM2025)Graduate School of Information Science, University of Hyogo (20-21 Oct. 2025)
Organized by: Intelligent Media Lab.
Venue: 7-1-28 Minatojima-minamimachi, Chuo-ku, Kobe, Hyogo 650-0047 (access details) |
Introduction
Welcome to the workshop on trustworthy foundation models 2025. This is an academic research event dedicated to exploring the transformative impact of these models on society, life, and research. As foundation models continue to integrate into various aspects of our daily lives, numerous trust issues have emerged, including but limited to concerns related to privacy, security, fairness, explainability, hallucination, bias, etc. This workshop aims to foster a collaborative environment by inviting esteemed researchers from both academia and industry to discuss the latest developments and ongoing challenges in creating trustworthy foundation models.
Invited Speakers
|
Prof. Hiroaki Kawashima, Graduate School of Information Science, University of Hyogo, JP
Prof. Kawashim was an assistant professor, senior lecturer, and associte professor at the Graduate School of Informatics, Kyoto University. Now, he is a professor as the Graduate School of Information Science, University of Hyogo. He was a visting scholar at the School of Electrical and Computer Engineering, Georgia Institute of Technology from 2010 to 2012 (JSPS Overseas research fellow) and a JST PRESTO researcher from 2014 to 2018.
|
|
Dr. Naoto Yanai, Principal Engineer, Technology Division, Panasonic Holding Corporation, JP
Dr. Yanai got the PhD degree in theoretical cryptography in 2014. He then joined Osaka University as an assistant professor in 2014 and became an associate professor in 2021. He has also researched in various topics in information security and has also published several top citations papers in blockchains and AI security. He has also organized an international workshop for AI security, SECAI from 2023. From 2024, he moved to Panasonic Holdings Corporation as a principal engineer and has researched in both academic and industrial topics. .
|
Program
Day 1 (Monday 20 Oct. 2025)
[10:00 – 11:50] Session 1: Research Progress from China
Chair: Xiaochun Cao, Sun Yat-sen University, China
Chair: Xiaochun Cao, Sun Yat-sen University, China
Time |
Speaker |
Title |
10:00 - 10:05 |
Essam Rashed (UH) |
Welcome note |
10:05 - 10:10 |
Xiaochun Cao (SYS) |
Opening |
10:10 - 10:35 |
Hua Zhang (CAS) |
Towards Explainable Open World Object Detection |
10:35- 11:00 |
Li Shen (SYS) |
Fine-Tuning Large Language Models with Forward-only Optimizers |
11:00 - 11:25 |
Yuchen Jiang (SYS) |
Flexibility-coupled Blockchain Federated Learning Architecture with Dynamic Malicious-client Filter for Provacy Computing |
11:25 - 11:50 |
Pengwen Dai (SYS) |
Visual Content Understanding and Protection |
UH: University of Hyogo, JP
SYS: Sun Yat-sen University, CN
CAS: Chinese Academy of Science, CN
SYS: Sun Yat-sen University, CN
CAS: Chinese Academy of Science, CN
[11:50 - 13:00] Break
[13:00 – 15:45] Session 2: Research Progress from Japan
Chair: Yang Cao, Institute of Science Tokyo, Japan
Chair: Yang Cao, Institute of Science Tokyo, Japan
Time |
Speaker |
Title |
13:00 - 13:05 |
Yang Cao (ST) |
Opening |
13:05 - 13:45 |
Hiroaki Kawashima (UH) |
[Invited talk][online] Machine Learning and Control of Collective Behavior |
13:45 - 14:25 |
Naoto Yanai (Pan) |
[Invited talk] Cybersecurity Research in LLM-Era |
14:25 - 14:45 |
Takao Murakami (ISM) |
[online] Augmented Shuffle Differential Privacy Protocols for Large-Domain Categorical and Key-Value Data |
14:45 - 15:05 |
Pengpeng Qiao (ST) |
Mixture of Experts in LLMs: An Overview and Our Federated Fine-Tuning Study |
15:05 - 15:25 |
Blaise Delattre (ST) |
Randomized Smoothing with Lipschitz Bounds: Extending Certified Robustness |
15:25 - 15:45 |
Yalda Zafari (QU) |
[online] Rethinking AI Models in Breast Cancer Care: Diagnosis, Prognosis, and Risk Assessments |
ST: Institute of Science Tokyo, JP
UH: University of Hyogo, JP
Pan: Panasonic Holding Corporation, JP
ISM: Institute of Statistical Mathematics, JP
QU: Qatar University, QA
UH: University of Hyogo, JP
Pan: Panasonic Holding Corporation, JP
ISM: Institute of Statistical Mathematics, JP
QU: Qatar University, QA
[15:45 - 16:15] Break
[16:15 – 17:15] Session 3: Panel Discussion
Topic: Recent advances in foundation models
Moderators: Masatoshi Yoshikawa, Osaka Seikei University, Japan & Hua Zhang, Chinese Academy of Sciences, China
Topic: Recent advances in foundation models
Moderators: Masatoshi Yoshikawa, Osaka Seikei University, Japan & Hua Zhang, Chinese Academy of Sciences, China
Panelists |
|
Xiaochun Cao, Sun Yat-sen University, China |
|
Yang Cao, Institute of Science Tokyo, Japan |
|
Naoto Yanai, Panasonic Holdings Corporation, Japan |
|
Blaise Delattre, Institute of Science Tokyo, Japan |
|
Ahmed Elboardy, University of Hyogo, Japan |
[17:15 – 18:15] Session 4: Flash presentations
Chairs: Takao Murakami, Institute of Statistical Mathematics, Japan & Li Shen, Sun Yat-sen University, China
Chairs: Takao Murakami, Institute of Statistical Mathematics, Japan & Li Shen, Sun Yat-sen University, China
Time |
Speaker |
Title |
17:15 - 17:30 |
Tomoyasu Okada (ST) |
Privacy Parameter Selection for Accuracy-First Differentially Private Fine-Tuning of Large Language Models |
17:30 - 17:45 |
Ahmed Darwish (NU) |
[online] LLMs Beyond Text: Regression Predictions with Regression-Aware Fine-Tuning (RAFT) |
17:45 - 18:00 |
Saher Mohamed (NU) |
[online] Uncertainty-aware medical image abnormality annotation |
18:00 - 18:15 |
Farid Tasharofi (FAU) |
FIND-Net: Fourier-Integrated Network with Dictionary Kernels for Metal Artifact Reduction |
ST: Institute of Science Tokyo, JP
NU: Nile University, EG
FAU: Friedrich-Alexander University, GE
NU: Nile University, EG
FAU: Friedrich-Alexander University, GE
[19:00 - 20:00] Social gathering
[20:00 – 21:00] Night Session
Chairs: TBA
Chairs: TBA
Day 2 (Tuesday 21 Oct. 2025)
[10:00 – 11:00] Excursion: Tour at Supercomputer Fugaku
Guide: Elisa Suka, University of Hyogo, Japan
Guide: Elisa Suka, University of Hyogo, Japan
[11:15 – 12:15] Session 5: Tutorial/Hands on
Moderators: Yiming Jia, University of Hyogo, Japan & Makoto Kanehira, University of Hyogo, Japan
Moderators: Yiming Jia, University of Hyogo, Japan & Makoto Kanehira, University of Hyogo, Japan
[12:15 - 13:00] Break
[13:00 – 14:30] Session 6: Student presentations
Chiars: Yuchen Jiang, Sun Yat-sen University, China & Pengpeng Qiao, Institute of Science Tokyo, Japan
Chiars: Yuchen Jiang, Sun Yat-sen University, China & Pengpeng Qiao, Institute of Science Tokyo, Japan
Time |
Speaker |
Title |
13:00 - 13:15 |
Li Lin (ST) |
How Robust Are Watermarks for Large Language Models? From Code Generation to RAG |
13:15 - 13:30 |
Hengyu Wu (ST) |
Backdoor Watermarking via Latent-Space Confusion Zones |
13:30 - 13:45 |
Ahmed Elboardy (UH) |
LLMs and VLMs in Healthcare: Beyond models architectures |
13:45 - 14:00 |
Hao Du (ST) |
Guarding Privacy in Fine-tuning LLMs: Attacks, Defenses, and DP Effectiveness |
14:00 - 14:15 |
Yiming Jia (UH) |
From Pixel to Precision: How Segmentation Enhances Diagnostic VLMs |
14:15 - 14:30 |
Meiyu Lin (ST) |
Backdoor Attack Meets SQL Injection: A Security Challenge for Text-to-SQL Models |
ST: Institute of Science Tokyo, JP
UH: University of Hyogo, JP
UH: University of Hyogo, JP
[14:30 – 14:40] Summary & closing
Chair: Essam Rashed, University of Hyogo, Japan
Chair: Essam Rashed, University of Hyogo, Japan
