AI coding agents still need human cleanup on nearly three out of four tasks in a real-world test
This story is only covered by news sources that have yet to be evaluated by the independent media monitoring agencies we use to assess the quality and reliability of news outlets on our platform. Learn more here.3 Articles
3 Articles
Researchers from Alibaba and several Asian universities have created a new coding agent test called Co-Design Bench from real user requests. Even the best model was able to fully solve only 26.4 percent of the tasks.
A benchmark based on actual requests shows that programming agents can solve much of a task, but they still leave incomplete assignments often: even the best evaluated model correctly completed only 26.4% of the tasks.
AI coding agents still need human cleanup on nearly three out of four tasks in a real-world test
Alibaba's new coding-agent benchmark, Co-Design Bench, tests AI agents on real user requests that already tripped them up in practice. Even GPT-6 Astra, the top performer, fully solves only 26.4 percent of tasks. On older benchmarks, the same models score above 99 percent. The article AI coding agents still need human cleanup on nearly three out of four tasks in a real-world test appeared first on The Decoder.
Coverage Details
Bias Distribution
- There is no tracked Bias information for the sources covering this story.
Factuality
To view factuality data please Upgrade to Premium
