Cross-platform-bench Collection The benchmarks evaluate LM agent on SWE/Computer-use tasks across different operating systems. • 2 items • Updated Jun 28