A Shanghai Jiao Tong University-led benchmark shows agents often find the right files but identify the exact bug-relevant lines only about 14% to 19% of the time.
What changed
The source article frames the development through the lens of AI Coding, AI benchmarks, Agents, Research, keeping the focus on concrete numbers, deployment constraints and the pressure they create for companies and policymakers.
Why it matters
The useful reading is not the launch headline alone. The bigger signal is how AI capability, cost, regulation and distribution channels are starting to determine which products can move from experiment to routine use.
Sources: CocoLoop, AI coding agents find the right file but miss the exact lines that matter, study shows（The Decoder）；SWE-Explore 基准研究（arXiv，上海交通大学等）