Lamini AI Partners with Meta to Enhance LLaMA's SQL Performance

Less than two weeks after the launch of Lamini Memory Tuning, Lamini AI has officially partnered with Meta.

Lamini AI announced Lamini Memory Tuning on June 13, wherein the tool has showcased the ability to improve factual accuracy while reducing hallucinations by as much as 95%.

“Lamini Memory Tuning is a research breakthrough that overcomes a seeming paradox in the AI world: achieving precise factual accuracy (i.e. no hallucinations) while upholding the generalisation capabilities that make LLMs valuable in the first place,” the startup said.

The tuning method was used on open-source models like LLaMa 3 and Mistral 3. Now, however, the company has partnered with Meta to improve LLaMa 3’s baseline performance by improving the quality of SQL queries.

As part of this, Meta published a repository of Llama 3 Lamini recipes to help tune Llama models, specifically for enterprises.

“Lamini Memory Tuning is a new tool you can use to embed facts into LLMs that improve factual accuracy and reduce hallucinations. Inspired by information retrieval, this method has set a new standard of accuracy for LLMs with less developer effort,” the repository stated.

According to Lamini, the memory-tuning tool is also able to reduce response times by 50% while also decreasing workloads for data teams and increasing the overall reliability of queries, subsequently increasing the accuracy rates as well.
This is not the first time a tool has attempted to improve the efficiency of SQL queries. Previously, researchers at Nanyang Technological University, Singapore University of Technology and Design, and Alibaba‘s DAMO Academy recently introduced LLM-R2, which was a query rewrite system that helped significantly boost SQL query efficiency.