# Hierarchical Belief Modeling for Zero-Shot Opponent Adaptation in Partially Observable Multi-Agent Navigation

arXiv: https://arxiv.org/abs/2609.12422 · HTML: https://mlsanity.com/paper/2609.12422/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 6 of its 14 references that are in the map.
- First posted on arXiv 11 Sep 2026.

## Builds on (most cited first, 6 of 6)

- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [E(n) Equivariant Graph Neural Networks](https://mlsanity.com/paper/2102.09844/) (2021)
- [Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning](https://mlsanity.com/paper/1803.11347/) (2018)
- [Meta Reinforcement Learning with Latent Variable Gaussian Processes](https://mlsanity.com/paper/1803.07551/) (2018)
- [Architecture Matters More Than Scale: A Comparative Study of Retrieval and Memory Augmentation for Financial QA Under SME Compute Constraints](https://mlsanity.com/paper/2604.17979/) (2026)

## Abstract (arXiv)

Lux AI Season 3 requires agents to act under partial observability, randomized episode level dynamics, and a best of five match structure that rewards both tactical execution and fast adaptation. We present HORIZON, a hierarchical agent that combines symmetry aware spatial perception, dual memory belief tracking, relic centric graph attention, information gain driven exploration, and an opponent conditioned policy mixture. HORIZON separates short horizon control from cross match meta reasoning, while auxiliary belief and world model objectives stabilize learning. Trained with PPO in a large scale JAX simulator, the resulting agent explicitly infers hidden game parameters and opponent style. Experiments show consistent gains in match win rate, episode win rate, adaptation gain, and league rating over strong recurrent and feed forward baselines.