IPAB Workshop - 29/01/2026 Speaker: Eric Liu Title: Three-Layer Model for Phase Transitions in Social Media Networks Abstract: In network science, capturing the non-linear and bursty nature of information diffusion is crucial, particularly within social media where user behaviour is influenced by both cognitive limits and structural echo chambers. This project focuses on establishing a theoretical framework for a three-layer dynamic model to decode the complex feedback loops driving viral cascades. This approach integrates dynamics across three scales: the micro-level, controlled by individual network exposure and cognitive boundaries; the meso-level, driven by local structural reinforcement; and the macro-level, the macro-level, where system-wide phase transitions can occur. The core objective is to understand the coupling mechanism where micro-level behaviours aggregate to trigger macroscopic criticality, which in turn acts as a top-down feedback field influencing individual activation probabilities at subsequent time steps. Speaker: Elle Miller Title: Realising tactile dexterity through world models Abstract: Vision-only robots fail the moment their line of sight is blocked. To achieve sophisticated manipulation, robots must learn to "feel." However, combining Reinforcement Learning (RL) with tactile sensing has remained a persistent challenge. In this talk, I show how integrating self-supervised world models with RL enables superhuman "blind" dexterity in simulation. I will also outline my vision for scaling these methods to enable physically-assistive caregiving robots. Jan 29 2026 13.00 - 14.00 IPAB Workshop - 29/01/2026 Eric Liu & Elle Miller MF2
IPAB Workshop - 29/01/2026 Speaker: Eric Liu Title: Three-Layer Model for Phase Transitions in Social Media Networks Abstract: In network science, capturing the non-linear and bursty nature of information diffusion is crucial, particularly within social media where user behaviour is influenced by both cognitive limits and structural echo chambers. This project focuses on establishing a theoretical framework for a three-layer dynamic model to decode the complex feedback loops driving viral cascades. This approach integrates dynamics across three scales: the micro-level, controlled by individual network exposure and cognitive boundaries; the meso-level, driven by local structural reinforcement; and the macro-level, the macro-level, where system-wide phase transitions can occur. The core objective is to understand the coupling mechanism where micro-level behaviours aggregate to trigger macroscopic criticality, which in turn acts as a top-down feedback field influencing individual activation probabilities at subsequent time steps. Speaker: Elle Miller Title: Realising tactile dexterity through world models Abstract: Vision-only robots fail the moment their line of sight is blocked. To achieve sophisticated manipulation, robots must learn to "feel." However, combining Reinforcement Learning (RL) with tactile sensing has remained a persistent challenge. In this talk, I show how integrating self-supervised world models with RL enables superhuman "blind" dexterity in simulation. I will also outline my vision for scaling these methods to enable physically-assistive caregiving robots. Jan 29 2026 13.00 - 14.00 IPAB Workshop - 29/01/2026 Eric Liu & Elle Miller MF2
