3DCodeEdit benchmark: models average 37% on 3D code edits
Reviewer-model feedback on renders helps more than extra reasoning or an agent harness, and can lift a 27B model past a 2.8T model's single-shot rate.
2 stories, newest first · one card per story
Reviewer-model feedback on renders helps more than extra reasoning or an agent harness, and can lift a 27B model past a 2.8T model's single-shot rate.
Empirical study finds direct whole-file generation beats iterative diff-based editing for Flutter/Dart code models on about 1,790 held-out tasks.
Today 1.6K readers · 2.4K views · 774 reads · all time 189K views