DAVIS: Planning Agent with Knowledge Graph-Powered Inner Monologue

ACL ARR 2025 February Submission5222 Authors

Abstract: Designing a generalist scientific agent capable of performing tasks in laboratory settings to assist researchers has become a key goal in recent Artificial Intelligence (AI) research. Unlike everyday tasks, scientific tasks are inherently more delicate and complex, requiring agents to possess a higher level of reasoning ability, structured and temporal understanding of their environment, and a strong emphasis on safety. Existing approaches often fail to address these multifaceted requirements. To tackle these challenges, we present DAVIS. Unlike traditional retrieval-augmented generation (RAG) approaches, DAVIS incorporates structured and temporal memory, which enables model-based planning. Additionally, DAVIS implements an agentic, multi-turn retrieval system, similar to human's inner monologue, allowing for a greater degree of reasoning over past experiences. Through internal planning before each step, DAVIS significantly reduces the likelihood of taking unsafe actions compared to baseline models. DAVIS demonstrates significant performance on the ScienceWorld benchmark, outperform previous approaches on 8 out of 9 elementary science subjects. In addition, DAVIS's World Model demonstrates competitive performance on the famous HotpotQA dataset for multi-hop question answering. To the best of our knowledge, DAVIS is the first RAG agent to employ an interactive retrieval method in RAG pipeline.

Paper Type: Long

Research Area: Question Answering

Research Area Keywords: knowledge graph, question answering, reinforcement learning, reflection, chain-of-thought, retrieval-augmented generation, knowledge-augmented methods, named entity recognition, relation extraction, reasoning, knowledge base QA

Contribution Types: Model analysis & interpretability, Publicly available software and/or pre-trained models

Languages Studied: English

Submission Number: 5222