🤖 AI Summary
To address hallucination, poor long-tail knowledge acquisition, and memory limitations of monolithic large language models (LLMs) in knowledge-intensive tasks, this paper proposes SMART, a multi-agent framework comprising four specialized agents that collaboratively execute subtask trajectories while integrating external knowledge retrieval and dynamic fusion mechanisms. SMART introduces the novel “long- and short-trajectory joint training” paradigm, enabling unified inter-agent coordination optimization and fine-grained action control—thereby overcoming key bottlenecks of conventional knowledge internalization or augmentation approaches in hallucination suppression and rare-knowledge acquisition. Evaluated on five knowledge-intensive benchmarks, SMART significantly outperforms state-of-the-art methods: it reduces average hallucination rate by 32.7%, improves factual accuracy by 18.4%, and substantially enhances task completion robustness.
📝 Abstract
Recent advancements in Large Language Models (LLMs) have led to significant breakthroughs in various natural language processing tasks. However, generating factually consistent responses in knowledge-intensive scenarios remains a challenge due to issues such as hallucination, difficulty in acquiring long-tailed knowledge, and limited memory expansion. This paper introduces SMART, a novel multi-agent framework that leverages external knowledge to enhance the interpretability and factual consistency of LLM-generated responses. SMART comprises four specialized agents, each performing a specific sub-trajectory action to navigate complex knowledge-intensive tasks. We propose a multi-agent co-training paradigm, Long-Short Trajectory Learning, which ensures synergistic collaboration among agents while maintaining fine-grained execution by each agent. Extensive experiments on five knowledge-intensive tasks demonstrate SMART's superior performance compared to widely adopted knowledge internalization and knowledge enhancement methods. Our framework can extend beyond knowledge-intensive tasks to more complex scenarios. Our code is available at https://github.com/yueshengbin/SMART.