MA-JEPA is a stochastic world model designed for multi‑agent reinforcement learning that replaces observation reconstruction with joint‑embedding prediction. It uses a categorical latent state and a causal Transformer trained on posterior and action‑conditioned dynamics to generate imagined trajectories for actor‑critic learning. The model achieves strong performance on the SMAC benchmark, matching or surpassing the best reported mean win rates on half of the evaluated maps.
arXiv:2501. 14622v5 Announce Type: replace Abstract: Learning efficient representations for decision-making policies is a challenge in imitation learning (IL).
The paper introduces CASTLE, a framework for fully decentralized multi‑agent reinforcement learning that uses two offline‑trained world models to provide in‑context guidance to independent agents. A Local Dynamics World Model captures each agent’s trajectory dynamics, while a Semantic‑Social World Model predicts short‑horizon task and social outcomes for candidate actions based on counterfactual simulator rollouts. During online learning, these frozen models supply prediction logits to an independent PPO policy, enabling agents to compare potential actions prospectively rather than relying solely on scalar rewards. CASTLE outperforms existing baselines on several multi‑particle benchmark tasks, achieving the highest mean final scores across 30 seeds.
JEPA‑TTT is a method that continuously adapts the latent dynamics predictor of a pretrained Joint‑Embedding Predictive Architecture (JEPA) world model during test time. It performs self‑supervised updates across episodes while keeping the visual encoder and reward head fixed, using dense replay to sample prediction windows from a growing buffer. In experiments on eight dynamics shifts across four continuous‑control environments, JEPA‑TTT reduces latent prediction error by 83% and improves planning performance by 153% compared to the frozen model.
arXiv:2608. 04934v1 Announce Type: cross Abstract: Training LLM agents commonly relies on supervised fine-tuning from expert trajectories or online reinforcement learning over human-specified tasks with handcrafted verifiers.
arXiv:2606. 30072v1 Announce Type: new Abstract: Cooperative tasks in Multi-Agent Reinforcement Learning (MARL) require agents to collectively maximize a shared return.
arXiv:2608.30067v1 Announce Type: cross Abstract: How do LLM agents come to both understand environments they act in and master tasks set within them? Through controlled experiments combining world-m...
Training LLM agents commonly relies on supervised fine-tuning from expert trajectories or online reinforcement learning over human-specified tasks with handcrafted verifiers. Though effective, both remain bottlenecked by externally specified tasks and supervision signals, limiting the scalability and diversity of agent training.
Recent studies on world modeling for Large Language Model (LLM) agents typically formulate the learning objective as next-observation prediction. However, this objective ties supervision to what a transition happens to reveal, which may omit the dynamics most relevant to the agent's current decision.
arXiv:2606. 24601v1 Announce Type: new Abstract: Multi-agent reinforcement learning (MARL) addresses the problem of training multiple agents that pursue collaborative, competitive, or mixed objectives.
World models enable agents to reason about future outcomes and learn policies from their knowledge of state transition, but existing approaches primarily focus on reconstructing future observations or...
arXiv:2608. 05989v1 Announce Type: new Abstract: Sample-efficient policy learning from pixels is a long-standing challenge in reinforcement learning (RL).
The paper introduces a Latent World Model (LWM) for robot navigation that predicts action‑conditioned latent feature compatibility instead of reconstructing future observations. By exploiting the correlation between spatial proximity and latent feature similarity, the model evaluates action consequences directly in latent space and supports counterfactual training using sampled action sequences. The learned world model can supervise policy learning from unlabeled video and further improve policies via reinforcement learning entirely within the model, eliminating the need for action annotations and additional environment interaction.