Applied Compute improves Qwen model efficiency on SWE-bench

Applied Compute increased the submit tool call rate from 22% to 60% for a Qwen thinking model on SWE-bench. The team achieved this by conditioning the rollout on an old production trace, reducing the turn count from 80 to 40 while maintaining the test pass rate.
Featured · Samuel Denton
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Morgan Stanley's Weaver warns of AI-compute bottleneck risks
- Claude Code 2.1.229 is about to be released
- Immich manages self-hosted photo libraries with AI features
- Open-source CUDA alternative targets portable AMD GPU code
- Silicon Data raises $30.5M to benchmark AI compute