ForeDreamer Dual-Agent Architecture for Open-Web Event Prediction

ForeDreamer uses a dual-agent framework to separate factual memory from experiential memory for open-web forecasting. A memory-processing subagent transforms noisy web evidence into structured, question-specific states to improve reasoning over raw retrieval results.