MAGE: Meta-Reinforcement Learning for Language Agents toward Strategic Exploration and Exploitation

Yang, Lu; Xu, Zelai; Xie, Minyang; Gao, Jiaxuan; Shok, Zhao; Wang, Yu; Wu, Yi

Computing Research RepositoryMar 4, 2026

MAGE: Meta-Reinforcement Learning for Language Agents toward Strategic Exploration and Exploitation - Part 1

MAGE: Meta-Reinforcement Learning for Language Agents toward Strategic Exploration and Exploitation - Part 2

MAGE: Meta-Reinforcement Learning for Language Agents toward Strategic Exploration and Exploitation - Part 3

MAGE: Meta-Reinforcement Learning for Language Agents toward Strategic Exploration and Exploitation - Part 4

Keep reading with a free trial. Try 1 month for $0

Meet DeepDyve

Get unlimited, instant access to over 150 million full-text scientific articles for less than the price of buying a single PDF.

150M+Research Papers

35M+Full-Text Streaming Papers

25M+Open Access Papers

$2/dayStarting Price

DeepDyve Collection

Access top-tier journals without the top-tier price. We partner directly with the world’s leading academic publishers to bring you legally compliant, instant access. Read the latest findings from Springer, Wiley, Oxford University Press, and more—all in one subscription.

Browse our collection of titles from leading publishers

AI Research Assistant

Accelerate your research with our advanced AI tools. Chat directly with your uploaded papers, and DeepDyve’s 50M+ LitStream and OA collections of full-text papers. AI responses come directly from the literature, footnoted and sourced, to provide the most authoritative and verifiable responses. AI outputs powered by OpenAI and Anthropic Claude. AI Smart Folders (coming soon)

Stop wasting time on paywalls, pirated sites and scattered journal portals

Streamline your research workflow with DeepDyve's comprehensive literature management solution for less than the price of buying a single PDF.

Try 1 month for $0