Blogs
Introducing OfficeQA Pro V2: A New Benchmark for Enterprise Grounded-Reasoning
A new benchmark for evaluating enterprise grounded reasoning on realistic, document-heavy tasks.
Databricks3x Faster Search: Parallel Test-Time Scaling with Instructed-Retriever-1
How parallel test-time scaling cuts search latency by more than 3× and answer latency by 2× without sacrificing quality.
DatabricksMemEx: A Programmable Scratchpad for LLM Agents
A programmable Python scratchpad that lets LLM agents use code as action to improve accuracy and reduce cost.
DatabricksMeet KARL: A Faster Agent for Enterprise Knowledge, Powered by Custom RL
Meet Databricks' knowledge agent, built with custom reinforcement learning for complex enterprise tasks.
Databricks