iiinova — where your digital self keeps evolving

Raul Steleac

@raulsteleac.bsky.social

36 followers 211 following 6 posts

PhD Student @edinburgh-uni.bsky.social; studying temporally extended behaviours in both single and multi-agent RL

Posts Replies Media

Raul Steleac @raulsteleac.bsky.social · 20/04/2026

Work done under the supervision of Mohan Sridharan and @dabelcs.bsky.social (many thanks)! If you’re at #ICLR2026 in Rio and want to chat, come find me during Poster Session 2! 🔥

020

Raul Steleac @raulsteleac.bsky.social · 20/04/2026

Finally, we use this multi-dimensional n-distance as a state representation for eigenoption discovery, leading to coordinated alignment patterns that are effective in aiding teams of agents in multiple downstream tasks. Also works with heterogeneous agent state spaces.

110

Raul Steleac @raulsteleac.bsky.social · 20/04/2026

Having a scalar describe the (mis)alignment of multiple agents is limiting, as it conflates the impact of different state features. We thus propose an information theoretic objective that disentangles the n-distance approximation across each individual state feature.

110

Raul Steleac @raulsteleac.bsky.social · 20/04/2026

We first generalise pairwise state distances to involve n elements by introducing n-distances in MDPs via our Fermat inter-agent distance. To define this metric, we introduce the Fermat state, a fictitious state of maximal alignment accross agents, and show how to approximate it.

110

Raul Steleac @raulsteleac.bsky.social · 20/04/2026

To discover such joint behaviours (options) in an unsupervised manner, we follow the intuition that: in the absence of an explicit objective, a natural basis for coordination among a group of agents can be achieved through patterns of alignment of their states. 🤔

110

Raul Steleac @raulsteleac.bsky.social · 20/04/2026

Really excited to present our recent work at #ICLR2026 this week! 🚨 We discover highly coordinated joint behaviours and integrate them into the skill sets of MARL agents, accelerating the search for effective joint strategies in downstream tasks. 🧵👇 Paper link: raulsteleac.github.io/iaro

1113