CausalEvolve: Towards Open-Ended Discovery with Causal Scratchpad

2026-03-15

AI Summary

Current evolutionary AI scientist agents suffer from low efficiency and performance instability in solving open-ended scientific problems due to the absence of targeted guidance and mechanisms for knowledge reuse. This work proposes the Causal Notebook framework, which integrates large language model–driven causal and abductive reasoning to dynamically identify key factors influencing evolution and uncover unexpected patterns. By doing so, it enables continuous, directed guidance of the evolutionary trajectory and fosters knowledge-driven innovation. Evaluated on four challenging open-ended scientific tasks, the approach significantly enhances evolutionary efficiency, stabilizes the optimization process, and discovers superior solutions compared to existing methods.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Search and Optimization: Evolutionary Computation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Models for Web evolution
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Evolve-based agent such as AlphaEvolve is one of the notable successes in using Large Language Models (LLMs) to build AI Scientists. These agents tackle open-ended scientific problems by iteratively improving and evolving programs, leveraging the prior knowledge and reasoning capabilities of LLMs. Despite the success, existing evolve-based agents lack targeted guidance for evolution and effective mechanisms for organizing and utilizing knowledge acquired from past evolutionary experience. Consequently, they suffer from decreasing evolution efficiency and exhibit oscillatory behavior when approaching known performance boundaries. To mitigate the gap, we develop CausalEvolve, equipped with a causal scratchpad that leverages LLMs to identify and reason about guiding factors for evolution. At the beginning, CausalEvolve first identifies outcome-level factors that offer complementary inspirations in improving the target objective. During the evolution, CausalEvolve also inspects surprise patterns during the evolution and abductive reasoning to hypothesize new factors, which in turn offer novel directions. Through comprehensive experiments, we show that CausalEvolve effectively improves the evolutionary efficiency and discovers better solutions in 4 challenging open-ended scientific tasks.

Problem

• evolutionary efficiency
• open-ended scientific discovery
• knowledge utilization
• oscillatory behavior
• evolution guidance

Innovation

• Causal Scratchpad
• Evolutionary Guidance
• Abductive Reasoning
• Open-Ended Discovery
• LLM-based Scientific Agent