Extract from Isha Marathe’s article “Paxton AI Launches A ‘Confidence Indicator’, Seeking to Bring Transparency to AI Accuracy”
Legal research startup Paxton AI announced on Tuesday the release of a new feature, called the Confidence Indicator, which will accompany its flagship research solution.
Additionally, the company also announced that it had received a nearly 94% accuracy score on the Stanford Legal Hallucination Benchmark, which analyzes how a tool handles legal tasks—such as statutory interpretation, case law analysis, and procedural questions—to score the output’s and accuracy.
Paxton AI has released the details of its scoring here. Additionally, the datasets that the tool is trained on and benchmarked off of is available on GitHub.
The Confidence Indicator release comes just weeks after Paxton AI announced a new tool called AI Citator, which works to create an index of legal materials that cite to a specific source while generating legal research answers.
What It Is: The ‘Confidence Indicator’ is a feature which rates every answer that Paxton AI’s legal research solution produces with a certain “confidence level.” These levels range from low, medium and high. Each confidence rating comes with a recommendation for the next steps the user can take.
