DeepSolution: Boosting Complex Engineering Solution Design via Tree-based Exploration and Bi-point Thinking
Zhuoqun Li, Haiyang Yu, Xuanang Chen, Hongyu Lin, Yaojie Lu, Fei Huang, Xianpei Han, Yongbin Li, Le Sun
2025-03-03
Summary
This paper talks about a new AI system called DeepSolution that helps solve complex engineering problems. It introduces two main things: SolutionBench, which is a way to test how good AI systems are at solving engineering problems, and SolutionRAG, which is a clever new method for coming up with solutions.
What's the problem?
Engineers often face really complicated problems that are hard to solve, even for smart computers. The current ways of using AI to help with these problems aren't good enough, especially when it comes to designing solutions that actually work in the real world with all its complicated rules and limits.
What's the solution?
The researchers created SolutionRAG, which uses two main tricks to solve engineering problems better. First, it explores different ideas like branches on a tree, looking at lots of possible solutions. Second, it uses something called 'bi-point thinking,' which helps it understand the problem from different angles. They also made SolutionBench, which is like a tough exam for AI systems to see how well they can solve real engineering problems.
Why it matters?
This matters because it could make solving complex engineering problems much easier and faster. If AI can come up with better solutions to tricky engineering challenges, it could lead to big improvements in things like building safer bridges, designing more efficient machines, or creating new technologies. It's a step towards having really smart computer assistants that can help engineers tackle the world's toughest problems.
Abstract
Designing solutions for complex engineering challenges is crucial in human production activities. However, previous research in the retrieval-augmented generation (RAG) field has not sufficiently addressed tasks related to the design of complex engineering solutions. To fill this gap, we introduce a new benchmark, SolutionBench, to evaluate a system's ability to generate complete and feasible solutions for engineering problems with multiple complex constraints. To further advance the design of complex engineering solutions, we propose a novel system, SolutionRAG, that leverages the tree-based exploration and bi-point thinking mechanism to generate reliable solutions. Extensive experimental results demonstrate that SolutionRAG achieves state-of-the-art (SOTA) performance on the SolutionBench, highlighting its potential to enhance the automation and reliability of complex engineering solution design in real-world applications.