#multi-hop question answering

TableRAG is a novel hybrid AI framework that integrates SQL and text retrieval to improve multi-hop question answering over documents containing both tables and text, demonstrating superior performance on complex datasets.

Microsoft introduces Phi-4-mini-Flash-Reasoning, a compact 3.8B parameter model optimized for efficient long-context reasoning and fast inference, outperforming previous models on complex benchmarks.