Well, the SWE-bench results provide some evidence. In just one year, the percentage of coding problems solved on the GitHub dataset (complex problem) has increased from 4.8% to 55%. impressive ? Indeed. source: www.swebench.com/viewer.html
Well, the SWE-bench results provide some evidence. In just one year, the percentage of coding problems solved on the GitHub dataset (complex problem) has increased from 4.8% to 55%. impressive ? Indeed. source: www.swebench.com/viewer.html