Search papers, labs, and topics across Lattice.
The paper introduces Agent2World, a multi-agent framework for generating symbolic world models by leveraging web searching, model implementation, and adaptive unit testing. This framework grounds world model generation in multi-agent feedback, addressing the limitations of static validation methods. Fine-tuning the model with trajectories generated by the interactive testing environment leads to a substantial improvement in world-model generation, achieving a 30.95% relative gain.
LLMs can learn to generate high-quality symbolic world models by interacting with a multi-agent system that provides adaptive, behavior-aware feedback, closing the gap between static validation and interactive execution.
Symbolic world models (e.g., PDDL domains or executable simulators) are central to model-based planning, but training LLMs to generate such world models is limited by the lack of large-scale verifiable supervision. Current approaches rely primarily on static validation methods that fail to catch behavior-level errors arising from interactive execution. In this paper, we propose Agent2World, a tool-augmented multi-agent framework that achieves strong inference-time world-model generation and also serves as a data engine for supervised fine-tuning, by grounding generation in multi-agent feedback. Agent2World follows a three-stage pipeline: (i) A Deep Researcher agent performs knowledge synthesis by web searching to address specification gaps; (ii) A Model Developer agent implements executable world models; And (iii) a specialized Testing Team conducts adaptive unit testing and simulation-based validation. Agent2World demonstrates superior inference-time performance across three benchmarks spanning both Planning Domain Definition Language (PDDL) and executable code representations, achieving consistent state-of-the-art results. Beyond inference, Testing Team serves as an interactive environment for the Model Developer, providing behavior-aware adaptive feedback that yields multi-turn training trajectories. The model fine-tuned on these trajectories substantially improves world-model generation, yielding an average relative gain of 30.95% over the same model before training. Project page: https://agent2world.github.io.
