Two research outcomes from our laboratory have been accepted for presentation at WI-IAT 2026 (The 25th IEEE/WIC International Conference on Web Intelligence and Intelligent Agent Technology). WI-IAT is recognised as one of the major international conferences on web intelligence and agent technology.

One presentation reports results on improving the accuracy of route planning for multiple vessels using the distributional reinforcement learning method IQN (Implicit Quantile Networks). The other proposes an efficient training method for reinforcement learning using LLMs. Both papers were accepted as Full Papers. Details are provided below. If you are interested and will be attending the conference, we would be grateful if you could join the presentations.

Presentation title: State-Dependent Risk Adaptation via Distributional Reinforcement Learning for Decentralized Multi-ASV Navigation
Authors: Tenyu Matsumoto, Donghui Lin, Manabu Ohta, and Fumito Uwano
Presentation date and time: TBD

Presentation title: Efficient LLM-Assisted Reinforcement Learning via Entropy-Driven Invocation Scheduling and Performance-Based Termination
Authors: Ryotaro Murakami, Donghui Lin, Manabu Ohta, and Fumito Uwano
Presentation date and time: TBD

Venue: Rydges South Bank Brisbane, Brisbane, Australia
Conference: The 25th IEEE/WIC International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT 2026)