Agent Lightning v1.0: Towards Harnessed Agentic RL
Open paper
Benchmark
Model
Rank
Results
code-generation-on-swe-bench-verified
Agent Lightning v1.0 (Qwen3.5-9B) (mini-SWE-agent) (extra data)
#71
Accuracy: 56.4
Rank counts only results with a code link.