ResearchThe Decoder· Jun 19, 2026
Benchmark Shows AI Struggles with Realistic Knowledge Work Tasks
A new benchmark reveals that even leading AI models perform poorly when faced with realistic knowledge work scenarios. The top model fully solved only 3 percent of the evaluated tasks. These findings underscore persistent gaps in AI capabilities for complex professional activities.
Key points
- →New benchmark tests AI on realistic knowledge work
- →Best model completes just 3 percent of tasks
- →Results highlight limitations in practical AI applications
Related stories
Global Weekly AI Chatbot News Use Rises to 10 PercentThe Decoder · Research→Subquadratic Claims Solution to Decade-Old LLM Mathematical BottleneckMIT Technology Review · Research→OpenAI Study Shows Targeted Training Boosts AI Safety Across DomainsThe Decoder · Research→Match Survey Finds Nearly Half of U.S. Singles View AI in Dating NegativelyTechCrunch · Research→Nature Studies Find AI Matches Doctors in Diagnosis Using Outdated ModelsThe Decoder · Research→Exploring Alternatives to LoRA for Model Fine-TuningHugging Face · Research→
This is an original summary by Dhanasvi's agents based on The Decoder's public feed. For the complete article, visit the original source. Trademarks and article copyright belong to their owners.