Autonomous Scientific Discovery via Iterative Meta-Reflection

2026-07-01

AI Summary

Current autonomous scientific discovery systems are constrained by predefined objectives and narrow search spaces, lacking the capacity to integrate and reflect upon their own findings. This work proposes DiscoPER, a framework that leverages large language models to dynamically generate and execute code without preset goals, enabling exploration of multimodal data while ensuring scientific rigor through statistical validation. Its key innovation lies in a second-order metacognitive reflection mechanism: the system treats its own discoveries as empirical data to identify structural patterns and cognitive blind spots, thereby actively expanding its hypothesis space. Evaluated on iNatDisco—a newly introduced benchmark for ecological knowledge discovery—DiscoPER successfully reproduces 8 out of 9 known patterns, achieving a hypothesis support rate of 72.7%, significantly outperforming conventional causal discovery methods and LLM-guided baselines.

Abstract

Autonomous scientific discovery systems offer the potential to accelerate research by automating the process of hypothesis generation and validation. However, current systems operate within constrained search spaces or require predefined research questions, limiting their capacity for true open-ended inquiry. Furthermore, while they generate hypotheses iteratively, they largely lack the ability to explicitly synthesize their own accumulated findings to uncover complex, interconnected phenomena. We introduce DiscoPER, an autonomous large language model-powered framework that conducts open-ended research by dynamically generating and executing code to explore datasets without pre-specified research objectives. To ensure rigorous scientific validity, every proposed discovery must pass statistical testing. To overcome the limitations of isolated search, our framework introduces a second-order reasoning mechanism that periodically analyzes its own accumulated discoveries. By treating prior discoveries as empirical data, DiscoPER identifies structural patterns, confounds, and epistemic gaps, actively redirecting hypothesis exploration toward uncharted regions of the search space. The search space is further expanded by incorporating tool use, enabling the system to explore hypotheses beyond structured metadata by seamlessly processing and extracting useful information from multimodal sources like images. Evaluated on iNatDisco, a new multimodal ecological knowledge benchmark with pattern-level ground truth obtained from peer-reviewed literature, DiscoPER recovers 8 of 9 known patterns with a 72.7% hypothesis support rate, outperforming both classical causal discovery and LLM-guided baselines. Ablations show that DiscoPER scales with more data, and confirms the benefits of second-order meta-reflection.

Problem

• autonomous scientific discovery
• open-ended inquiry
• hypothesis generation
• meta-reflection
• search space limitation

Innovation

• autonomous scientific discovery
• second-order meta-reflection
• open-ended hypothesis generation
• multimodal data exploration
• LLM-powered framework