[DMind Benchmark, launched by the Minara team, officially accepted by KDD 2026, becomes the first large-scale evaluation benchmark for Digital Assets × AI to appear at an international top-tier conference] BlockBeats News, July 9 — The DMind Benchmark, introduced by the Minara team in 2025, has recently had its paper officially accepted by the Datasets & Benchmarks Track of the KDD 2026 main conference, marking the first peer-reviewed large-scale evaluation benchmark in the intersection of Digital Assets × AI to enter the main conference (not workshop or demo) of an international top-tier event. KDD is a premier international conference in the fields of data mining and machine learning, with an acceptance rate of approximately 29% for this track this year. The DMind Benchmark evaluates models' multi-step reasoning capabilities in real-world digital asset scenarios. It consists of 3,154 objective questions and 389 open-ended tasks, meticulously reviewed by domain experts, covering nine subfields including DeFi, Tokenomics, contract security, and more. It has systematically assessed 31 cutting-edge large models, including GPT, Claude, and Gemini. The dataset previously topped Hugging Face's site-wide Trending list, with over 13,000 cumulative downloads. Re-evaluation conducted in July 2026 revealed that despite advancements in six latest flagship models, the gap in domain depth remains unbridged. Leveraging this benchmark and DMind, efforts are underway to build a dedicated intelligent foundational layer for the digital asset domain, enabling AI to truly understand the world of digital assets. [Original link]