Why did Collide develop RIGGS, a purpose-built LLM for oil and gas?

Generalized AI often fails to provide consistent and accurate results for the oil and gas industry, even for basic terminology. Collide built RIGGS to address this by creating a large language model specifically designed and fine-tuned for the unique complexities of the oil and gas sector.

What is RIGGS and what foundational model was it built upon?

RIGGS is a purpose-built large language model developed by Collide for the oil and gas industry. It began as a proof-of-concept, fine-tuned using a small subset of domain-specific data on the Llama 3.1 8B parameter model.

How was RIGGS' initial performance benchmarked?

RIGGS' performance was benchmarked by its ability to pass a standardized petroleum engineering exam. This exam serves as Collide's internal measure for how well a model comprehends core oil and gas concepts.

What were the initial results of RIGGS' performance benchmark?

Initially, RIGGS scored 35% on the petroleum engineering exam, which was a significant improvement over Llama's baseline score of 18%. The Collide team is continuously working to enhance RIGGS' understanding and accuracy through further domain-specific training data.

Why We Built RIGGS: A Purpose-Built LLM for Oil & Gas