LLMs outperform RL in portfolio allocation by filtering aged financial news

Researchers have explored the impact of news aging on investment portfolio allocation using reinforcement learning (RL) and large language models (LLMs). A study involving historical tests from 2023-2025 found that while RL policies can respond to timing information, there was no consistent benefit from retaining aging news signals. For LLMs directly selecting portfolios, filtering out older news proved most effective, yielding an average annual net return of 18.43% when news delivery was delayed by four weeks, outperforming all RL strategies that incorporated news. AI

IMPACT Demonstrates a practical application of LLMs in financial decision-making, potentially improving investment strategies.

RANK_REASON Academic paper on applying AI methods to a specific domain. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• arXiv
• large-language models
• reinforcement learning

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →