Uncovering Untapped Potential in Sample-Efficient World Model Agents
This addresses sample efficiency problems for reinforcement learning researchers, though it appears incremental as it combines existing techniques in a novel way.
The paper tackled the limitations of token-based world models (TBWMs) by introducing Simulus, a modular agent that integrates multi-modality tokenization, intrinsic motivation, prioritized replay, and regression-as-classification, achieving state-of-the-art sample efficiency for planning-free WMs across three diverse benchmarks.
World model (WM) agents enable sample-efficient reinforcement learning by learning policies entirely from simulated experience. However, existing token-based world models (TBWMs) are limited to visual inputs and discrete actions, restricting their adoption and applicability. Moreover, although both intrinsic motivation and prioritized WM replay have shown promise in improving WM performance and generalization, they remain underexplored in this setting, particularly in combination. We introduce Simulus, a highly modular TBWM agent that integrates (1) a modular multi-modality tokenization framework, (2) intrinsic motivation, (3) prioritized WM replay, and (4) regression-as-classification for reward and return prediction. Simulus achieves state-of-the-art sample efficiency for planning-free WMs across three diverse benchmarks. Ablation studies reveal the individual contribution of each component while highlighting their synergy. Our code and model weights are publicly available at https://github.com/leor-c/Simulus.