Contextual AI unveiled its grounded language mannequin (GLM) at present, claiming it delivers the best factual accuracy within the business by outperforming main AI programs from Google, Anthropic and OpenAI on a key benchmark for truthfulness.
The startup, based by the pioneers of retrieval-augmented technology (RAG) know-how, reported that its GLM achieved an 88% factuality rating on the FACTS benchmark, in comparison with 84.6% for Google’s Gemini 2.0 Flash, 79.4% for Anthropic’s Claude 3.5 Sonnet and 78.8% for OpenAI’s GPT-4o.
Whereas giant language fashions have reworked enterprise software program, factual inaccuracies — typically known as hallucinations — stay a essential problem for enterprise adoption. Contextual AI goals to resolve this by making a mannequin particularly optimized for enterprise RAG purposes the place accuracy is paramount.
“We knew that a part of the answer could be a method known as RAG — retrieval-augmented technology,” stated Douwe Kiela, CEO and cofounder of Contextual AI, in an unique interview with VentureBeat. “And we knew that as a result of RAG is initially my concept. What this firm is about is basically about doing RAG the proper manner, to sort of the following degree of doing RAG.”
The corporate’s focus differs considerably from general-purpose fashions like ChatGPT or Claude, that are designed to deal with every part from inventive writing to technical documentation. Contextual AI as an alternative targets high-stakes enterprise environments the place factual precision outweighs inventive flexibility.
“You probably have a RAG downside and also you’re in an enterprise setting in a extremely regulated business, you haven’t any tolerance in anyway for hallucination,” defined Kiela. “The identical general-purpose language mannequin that’s helpful for the advertising and marketing division isn’t what you need in an enterprise setting the place you might be far more delicate to errors.”
How Contextual AI makes ‘groundedness’ the brand new gold normal for enterprise language fashions
The idea of “groundedness” — guaranteeing AI responses stick strictly to info explicitly supplied within the context — has emerged as a essential requirement for enterprise AI programs. In regulated industries like finance, healthcare and telecommunications, firms want AI that both delivers correct info or explicitly acknowledges when it doesn’t know one thing.
Kiela provided an instance of how this strict groundedness works: “Should you give a recipe or a formulation to a normal language mannequin, and someplace in it, you say, ‘however that is solely true for many circumstances,’ most language fashions are nonetheless simply going to provide the recipe assuming it’s true. However our language mannequin says, ‘Really, it solely says that that is true for many circumstances.’ It’s capturing this extra little bit of nuance.”
The power to say “I don’t know” is an important one for enterprise settings. “Which is known as a very highly effective function, if you concentrate on it in an enterprise setting,” Kiela added.
Contextual AI’s RAG 2.0: A extra built-in solution to course of firm info
Contextual AI’s platform is constructed on what it calls “RAG 2.0,” an method that strikes past merely connecting off-the-shelf elements.
“A typical RAG system makes use of a frozen off-the-shelf mannequin for embeddings, a vector database for retrieval, and a black-box language mannequin for technology, stitched collectively by means of prompting or an orchestration framework,” in accordance with an organization assertion. “This results in a ‘Frankenstein’s monster’ of generative AI: the person elements technically work, however the entire is way from optimum.”
As an alternative, Contextual AI collectively optimizes all elements of the system. “We have now this mixture-of-retrievers element, which is known as a solution to do clever retrieval,” Kiela defined. “It seems to be on the query, after which it thinks, basically, like a lot of the newest technology of fashions, it thinks, [and] first it plans a method for doing a retrieval.”
This whole system works in coordination with what Kiela calls “the very best re-ranker on this planet,” which helps prioritize essentially the most related info earlier than sending it to the grounded language mannequin.
Past plain textual content: Contextual AI now reads charts and connects to databases
Whereas the newly introduced GLM focuses on textual content technology, Contextual AI’s platform has not too long ago added assist for multimodal content material together with charts, diagrams and structured knowledge from common platforms like BigQuery, Snowflake, Redshift and Postgres.
“Probably the most difficult issues in enterprises are on the intersection of unstructured and structured knowledge,” Kiela famous. “What I’m largely enthusiastic about is basically this intersection of structured and unstructured knowledge. A lot of the actually thrilling issues in giant enterprises are smack bang on the intersection of structured and unstructured, the place you will have some database data, some transactions, perhaps some coverage paperwork, perhaps a bunch of different issues.”
The platform already helps quite a lot of complicated visualizations, together with circuit diagrams within the semiconductor business, in accordance with Kiela.
Contextual AI’s future plans: Creating extra dependable instruments for on a regular basis enterprise
Contextual AI plans to launch its specialised re-ranker element shortly after the GLM launch, adopted by expanded document-understanding capabilities. The corporate additionally has experimental options for extra agentic capabilities in improvement.
Based in 2023 by Kiela and Amanpreet Singh, who beforehand labored at Meta’s Basic AI Analysis (FAIR) staff and Hugging Face, Contextual AI has secured clients together with HSBC, Qualcomm and the Economist. The corporate positions itself as serving to enterprises lastly notice concrete returns on their AI investments.
“That is actually a chance for firms who’re perhaps underneath strain to start out delivering ROI from AI to start out taking a look at extra specialised options that really resolve their issues,” Kiela stated. “And a part of that basically is having a grounded language mannequin that’s perhaps a bit extra boring than a normal language mannequin, however it’s actually good at ensuring that it’s grounded within the context and which you could actually belief it to do its job.”
