Diffbot, renowned for its extensive web knowledge index, has launched a new AI model aimed at tackling the significant issue of factual accuracy. The model utilizes GraphRAG, an open-source system built on Meta’s LLama 3.3, but setting it apart by using real-time data from Diffbot’s ever-updating Knowledge Graph. CEO Mike Tung advocates separating the model from stored knowledge, emphasizing tool usage to query external information.
The Knowledge Graph, constantly refreshed with new data, enhances the model’s transparency and accuracy, allowing it to pull current information from the web rather than relying on outdated data. Diffbot’s approach significantly outperforms traditional models, achieving high scores on benchmarks such as FreshQA and MMLU-Pro. With its open-source release, the model promotes flexibility and data privacy for companies needing customized solutions.
Experts suggest its Knowledge Graph-centric design could be invaluable for enterprises prioritizing factual precision and auditability, with major corporations already utilizing Diffbot’s data services. This novel method offers an alternative to the prevalent trend of developing larger AI models, proposing instead a focus on effective knowledge organization and access.
The ainewsarticles.com article you just read is a brief synopsis; the original article can be found here: Read the Full Article…
