Technical Sharing Session on AI Safety

Rishabh is a PHD Candidate at the Singapore University of Technology and Design (SUTD), whose research primarily centers on making NLP models more robust. He is currently exploring ways to make LLMs more harmless (safe) while staying helpful (performant). As part of his safety research, his team released a safety benchmark Red-Eval, and an alignment approach Red-Instruct.

Rishabh has published over 20 research papers, and recent papers at top places (ACL, EMNLP, ICASSP) involve parameter-efficient learning in language models including dynamic prompts, hypernetworks, and adapters for speech. In the line of PE, he has also explored ways to prune adapters using hyper surface geometry.

At the beginning of his PhD, he worked on gender-debiasing algorithms in language models where they removed information from directions in the embedding space that encodes gender information. Another work studied the interpretability of Transformers, where they theoretically analysed if the self-attention is identifiable and proposed identifiable transformers (accepted at ACL 2021).

Rishabh also worked at Salesforce as an NLP research intern in making prompt-tuning dynamic (EMNLP 2022), and was a research intern at AWS AI at Amazon, California, in converting LLMs semi-parametric (ACL 2023).

Dr. Wenxuan Zhang is currently a research scientist at Alibaba DAMO Academy. He received his Ph.D. degree from The Chinese University of Hong Kong under the supervision of Professor Wai Lam, and then joined Alibaba Singapore with the Ali Star award.

He mainly works on natural language processing and trustworthy AI. His research aims to advance NLP models that are inclusive, supporting diverse languages and cultures, while also trustworthy by improving their safety and robustness.

He has published multiple papers as the first or corresponding author in top-tier AI conferences and journals, including ICLR, NeurIPS, ACL, EMNLP, SIGIR, WWW, TOIS, and TKDE. He also regularly serves on the program committees of multiple leading conferences and journals. He organized a tutorial at IJCAI 2023 and received the outstanding speaker award of MLNLP 2023.

Dr. Sunil Sivadas is Director of NEXT Gen Tech, a deep tech innovation and translation R&D team at NCS which he co-founded in 2018. He is part of NCS AI Strategy taskforce responsible for creating differentiating capabilities, assets and partnerships.

He was concurrently Deputy Director of the Singtel AI Lab at Nanyang Technological University, where he directed research agenda and managed teams working on projects across Generative AI, Computer Vision, Metaverse and Robotics. He has more than 20 years of experience developing machine learning algorithms and building AI systems.

Before joining NCS he was Unit Head of Human Language Technology department in I2R, A*STAR and Member of Technical Staff at Nokia-Bell labs, Finland. He is an inaugural NCS Distinguished Engineer and Adjunct Associate Professor in the School of Computer Science & Engineering, NTU.