Researchers introduce Analogical Deep Research (ADR), a new task and benchmark (ADR-bench) designed to test whether Large Language Model agents can effectively retrieve and integrate historical analogies for foresight analysis. The study identifies that current agents struggle because they match surface features rather than underlying causal mechanisms.
To address this, the authors propose Causal Analogical Researcher (CANA), an agentic framework that uses structural decomposition and feedback to align mechanisms across analogies. CANA achieves up to 10% improvement in historical analogy generation and surpasses state-of-the-art deep research agents on ADR-bench.
Case studies involving ongoing events confirm the effectiveness of CANA in leveraging historical analogies for accurate foresight analysis.