PseudoAct: Leveraging Pseudocode Synthesis for Flexible Planning and Action Control in Large Language Model Agents

Researchers introduce PseudoAct, a new framework that uses pseudocode synthesis to improve large language model agent planning and action control. The method achieves significant performance improvements over existing reactive approaches, with a 20.93% absolute gain in success rate on FEVER benchmark and new state-of-the-art results on HotpotQA.

• →PseudoAct addresses limitations of reactive decision-making in LLM agents by using structured pseudocode planning.
• →The framework reduces redundant tool usage and prevents infinite loops in complex long-horizon tasks.
• →Method achieved 20.93% absolute improvement in success rate on FEVER benchmark testing.
• →PseudoAct enables explicit control flow including sequencing, conditionals, loops, and parallel composition.
• →The approach sets new state-of-the-art performance on HotpotQA question-answering benchmark.