Brief

Researchers have developed M3, a system that uses conversational LLMs to simplify access and analysis of complex clinical databases like MIMIC-IV. M3 allows users to query the data using natural language, translating questions into SQL queries for execution. Evaluations showed high accuracy for models like Claude Sonnet 4 and the open-weights gpt-oss-20B, demonstrating the viability of local, privacy-preserving deployment for sensitive medical data. AI

Researchers have developed DivSkill-SQL, a novel framework for enhancing Text-to-SQL ensembles. This method optimizes complementary skills by training new agents on examples that the existing ensemble fails on, thereby increasing the probability of generating at least one correct SQL candidate. The framework demonstrated significant improvements, boosting accuracy by up to 11.1 points on Snowflake and 8.3 points on BigQuery when tested with Opus-4.6 and GPT-5.4 base models on the Spider2-Lite dataset. Notably, these optimized skills showed transferability across different SQL dialects and task formulations, with error analysis indicating a reduction in hallucinations and more reliable complementary skills. AI

Replit has launched a new, powerful search engine designed to help users find content within its platform in under 30 seconds. The engine indexes a wide range of items, including Repls, templates, code, users, and community content. This initiative addresses a significant user pain point, as 80% of users previously abandoned the search function due to its ineffectiveness. Replit built the search engine using Elasticsearch for indexing and Apache Spark for data pipelines, with plans to expand code search capabilities to all files in every Repl. AI