AI BriefingSup AI minimizes hallucination with multi-LLM ensemble
Key point
Sup AI launches, reducing hallucination by measuring entropy across 339 LLMs to synthesize answers.
Details
Sup AI has unveiled a technology that runs 339 LLMs in parallel to solve hallucination, a chronic problem in LLMs.
Rather than simply aggregating answers, it synthesizes answers by measuring the Entropy of each answer segment. Segments with high entropy are considered more likely to be hallucinations and are given lower weight, while segments with low entropy are judged to be more accurate and are given higher weight.
Through this approach, it recorded 52.15% on the 'Humanity's Last Exam' benchmark, which is 7.41 points higher than individual models.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Toss Bank Saves 300 Hours Annually with AI Consultation