Get Started
Home
Topics
Search
Library
Research questionHow can LLM agent configurations be jointly optimized when prompts, tools, and parameters interact?LLM agents depend on interacting prompts, tool descriptions, and runtime parameters, so changing one component can alter the effects of others. Manual refinement is slow, while optimizing components separately can miss configurations that work well together.
AI
AI Agents
Evaluation & Benchmarks
Machine Learning
Neural and Evolutionary Computing
Latest papersRecent research connected to this question, newest first.Evolving Excellence: Automated Optimization of LLM-based AgentsThe source describes a no-code evolutionary platform that takes a benchmark script and natural-language goals, identifies configurable components, extracts signals from execution logs, and evolves configurations without architectural changes. It was evaluated on ALE, Mini-SWE, CrewAI, and MathTales-Teacher agents across programming, code optimization, mathematical reasoning, accuracy, and token-use tasks, including commercial and smaller open-source models; the reported improvements are benchmark-specific.research paper · Sep 3, 2026
Related questions
How can we train and evaluate LLM agents for tool use across single- and multi-turn workflows with serial or parallel calls?How can LLM agents generalize to unseen tasks without directly fine-tuning their policies?How can long-horizon LLM agents preserve answer quality under tight prompt-token budgets?How can LLM agents reconcile conflicting user instructions, stored knowledge, and live tool observations before acting?