EvoDuet: Bilevel Co-Evolution of Web Searching and Task Solving for Scientific Discovery

2026-09-30

AI Summary

This study addresses the stagnation and redundant retrieval issues in evolutionary search with large language models caused by a lack of external knowledge. To this end, we propose a bi-level co-evolutionary framework operating under fixed parameters. Specifically, the method introduces a dynamic retrieval gating mechanism that assesses knowledge gaps to prevent redundant searches. The inner loop optimizes query ranking, while the outer loop generates candidate solutions in parallel and logs historical results, thereby enabling the co-evolution of solutions and queries. Experimental results demonstrate that our approach significantly improves discovery gains, outperforming state-of-the-art methods across eight tasks. Furthermore, its generalizability is validated across multiple evolutionary frameworks.

Abstract

Evolutionary search with large language models (LLMs) can stall when progress requires external knowledge the model lacks. Supplying relevant documents helps, but simply adding web search tool can keep returning the same pages as solutions change. We introduce EvoDuet, a bi-level optimization method that co-evolves solutions and search queries with fixed model parameters. At each iteration, a retrieval gate lets the LLM assess its knowledge gap and choose to retrieve new documents, reuse stored ones, or proceed without them. An inner loop refines queries and ranks documents by the solution scores they are predicted to yield; an outer loop generates candidates in parallel from these documents and records the evaluated outcomes for later searches. Across 21 optimization tasks with one candidate per iteration, EvoDuet raises OpenEvolve's normalized discovery gain from 74.1% to 78.0% with GPT-5.6-Luna and from 61.3% to 82.3% with Gemini-3.8-Flash, whereas Qwen3.5-9B does not benefit. Our best runs surpass the previously reported best scores on eight tasks, including Swap Reduction on Q20 and Rosetta, and match them on three more. EvoDuet also improves with other scaffolds (e.g., Top-K, EvoX) on Sums/Diffs and Denoising, demonstrating its applicability across evolutionary search scaffolds.

Problem

• Evolutionary search
• Large language models
• Web searching
• Scientific discovery
• Knowledge gap

Innovation

• Bilevel Co-Evolution
• Retrieval Gate
• Evolutionary Search
• Web Searching
• Scientific Discovery