Friday, 2nd August - 11am Hongru Wang : Seminar Title: Integrating Internal Reasoning and External Acting for Personalized Dialogue Agent Abstract: Large Language Models (LLMs) or Language Agents achieves exceptional performance at many downstream tasks and applications. However, the internal reasoning processing still lack explainability and transparency due to black-box nature, while external acting remains underexplore due to complex real-world interaction. In this talk, I will cover a range of methods to 1) enhance the explainability and transparency of reasoning (e.g., Chain-of-X, meta reasoning theory); and 2) to empower LLMs to call different tools / knowledge sources (e.g., RAG, multiple sources planning, tool learning). Moreover, I will discuss a unified view that integrates reasoning and acting from a tool perspective, regarding reasoning as a cognitive tool and acting as a physical tool. This perspective highlights the synergy between reasoning and action, emphasizing their complementary roles in the functionality of LLMs. Bio: Hongru WANG is a 3rd year Ph.D. candidate at The Chinese University of Hong Kong and currently visiting University of Edinburgh. His research interest includes dialogue system, tool learning and language agent. He has published more than 20 papers in top international conferences such as ACL, EMNLP, NAACL, COLING, etc, and some representative works include Cue-CoT, SAFARI, UniMS-RAG, SelfDC etc. He has won the Best Paper Award of the International Doctoral Student Forum, the WWW2024 Online Safety Prize Challenge Champion, and many other honors. His personal homepage is https://rulegreen.github.io/ Aug 02 2024 11.00 - 12.00 Friday, 2nd August - 11am Hongru Wang : Seminar This event is co-organised by ILCC and by the UKRI Centre for Doctoral Training in Natural Language Processing, https://nlp-cdt.ac.uk. IF G.07 and on Teams Contact
Friday, 2nd August - 11am Hongru Wang : Seminar Title: Integrating Internal Reasoning and External Acting for Personalized Dialogue Agent Abstract: Large Language Models (LLMs) or Language Agents achieves exceptional performance at many downstream tasks and applications. However, the internal reasoning processing still lack explainability and transparency due to black-box nature, while external acting remains underexplore due to complex real-world interaction. In this talk, I will cover a range of methods to 1) enhance the explainability and transparency of reasoning (e.g., Chain-of-X, meta reasoning theory); and 2) to empower LLMs to call different tools / knowledge sources (e.g., RAG, multiple sources planning, tool learning). Moreover, I will discuss a unified view that integrates reasoning and acting from a tool perspective, regarding reasoning as a cognitive tool and acting as a physical tool. This perspective highlights the synergy between reasoning and action, emphasizing their complementary roles in the functionality of LLMs. Bio: Hongru WANG is a 3rd year Ph.D. candidate at The Chinese University of Hong Kong and currently visiting University of Edinburgh. His research interest includes dialogue system, tool learning and language agent. He has published more than 20 papers in top international conferences such as ACL, EMNLP, NAACL, COLING, etc, and some representative works include Cue-CoT, SAFARI, UniMS-RAG, SelfDC etc. He has won the Best Paper Award of the International Doctoral Student Forum, the WWW2024 Online Safety Prize Challenge Champion, and many other honors. His personal homepage is https://rulegreen.github.io/ Aug 02 2024 11.00 - 12.00 Friday, 2nd August - 11am Hongru Wang : Seminar This event is co-organised by ILCC and by the UKRI Centre for Doctoral Training in Natural Language Processing, https://nlp-cdt.ac.uk. IF G.07 and on Teams Contact
Aug 02 2024 11.00 - 12.00 Friday, 2nd August - 11am Hongru Wang : Seminar This event is co-organised by ILCC and by the UKRI Centre for Doctoral Training in Natural Language Processing, https://nlp-cdt.ac.uk.
