SWE-Bench Pro Interactive Benchmark - a Anonym01048 Collection

Anonym01048 's Collections

updated Aug 30

Issue-solving and issue+QA configs over a shared 701-task SWE-Bench Pro subset.

Anonym01048/SWE-Bench-Pro-interactive-issue

Viewer • Updated Aug 30 • 2.1k • 325

Note Issue-solving prompt sequences — paper conditions Single (single-prompt), Split (split-prompt), Concat (split-prompt-concat) over the same 701 tasks.
Anonym01048/SWE-Bench-Pro-interactive-issue-qa

Viewer • Updated Aug 30 • 1.98k • 452

Note QA-augmented trajectories — Split+QA placements (first/last/random), QA-only, and the TFAS-optimized selection (qa-tfas-optimized-sol), with reference answers and golden proof scripts.