RWML：面向 LLM 智能体的强化世界模型学习

View PDF HTML (experimental)

Abstract:Large language models (LLMs) have achieved strong performance in language-centric tasks. However, in agentic settings, LLMs often struggle to anticipate action consequences and adapt to environment dynamics, highlighting the need for world-modeling capabilities in LLM-based agents. We propose Reinforcement World Model Learning (RWML), a self-supervised method that learns action-conditioned world models for LLM-based agents on textual states using sim-to-real gap rewards. Our method aligns simulated next states produced by the model with realized next states observed from the environment, encouraging consistency between internal world simulations and actual environment dynamics in a pre-trained embedding space. Unlike next-state token prediction, which prioritizes token-level fidelity (i.e., reproducing exact wording) over semantic equivalence and can lead to model collapse, our method provides a more robust training signal and is empirically less susceptible to reward hacking than LLM-as-a-judge. We evaluate our method on ALFWorld and $\tau^2$ Bench and observe significant gains over the base model, despite using no expert data, task-success rewards, or strong-LLM annotations - only an off-the-shelf embedding model as a distance metric. When combined with task-success rewards, our method outperforms direct task-success reward RL by 6.9 and 5.7 points on ALFWorld and $\tau^2$ Bench respectively; it further outperforms expert-data training on ALFWorld and is comparable to it on $\tau^2$ Bench.

Submission history

From: Xiao Yu [view email]
[v1] Thu, 5 Feb 2026 16:30:08 UTC (408 KB)
[v2] Mon, 9 Feb 2026 04:31:11 UTC (408 KB)
[v3] Sun, 1 Nov 2026 23:07:53 UTC (430 KB)