01 / AI · Claim vs. Check · Analysis
Google DeepMind benchmark tests professional AI tasks
The assessment covers complex visual analysis and the completion of professional work, including reports and spreadsheets.
Google DeepMind released a benchmark to assess AI agents on professional tasks, DigitalToday, a Korean technology news publication, reported on 16 Nov 2026. A benchmark is an assessment used to measure performance.
According to DigitalToday, the assessment covers work from analysing visual information to completing a finished document. Its scope extends beyond answering individual questions to carrying out professional tasks.
DigitalToday reported that Google DeepMind developed the Multimodal Knowledge Agent (MKA) benchmark with AI analysis firm Artificial Analysis. The assessment brings visual analysis and the production of work documents into the same evaluation.
Analysing complex drawings
According to DigitalToday, MKA includes architectural drawings and electronic circuit diagrams. These tasks assess an agent's ability to interpret complex visual information.
The report also identified clinical images among the materials assessed. The visual tasks therefore extend beyond technical drawings to images used in medicine.
Completing reports and spreadsheets
DigitalToday reported that the benchmark also measures the ability to complete reports and spreadsheets. A spreadsheet is a document that organises information in tables.
In the assessment described by DigitalToday, completing these documents is part of judging professional task performance. The scope carries the evaluation through from reading visual information to finishing a work document.