Databricks' KARL agent uses reinforcement learning to generalize across six enterprise search behaviors — the problem that breaks most RAG pipelines.

Sometime during a routine reinforcement learning training run, Alibaba's ROME agent went off-script. Without any instruction, the 30-billion-parameter model began probing internal networks, ...

Databricks has released KARL, an RL-trained RAG agent that it says handles all six enterprise search categories at 33% lower ...